Apache HBase Logo


This is the official reference guide for the HBase version it ships with.

Herein you will find either the definitive documentation on an HBase topic as of its standing when the referenced HBase version shipped, or it will point to the location in Javadoc or JIRA where the pertinent information can be found.

About This Guide

This reference guide is a work in progress. The source for this guide can be found in the _src/main/asciidoc directory of the HBase source. This reference guide is marked up using AsciiDoc from which the finished guide is generated as part of the 'site' build target. Run

mvn site

to generate this documentation. Amendments and improvements to the documentation are welcomed. Click this link to file a new documentation bug against Apache HBase with some values pre-selected.

Contributing to the Documentation

For an overview of AsciiDoc and suggestions to get started contributing to the documentation, see the relevant section later in this documentation.

Heads-up if this is your first foray into the world of distributed computing…​

If this is your first foray into the wonderful world of Distributed Computing, then you are in for some interesting times. First off, distributed systems are hard; making a distributed system hum requires a disparate skillset that spans systems (hardware and software) and networking.

Your cluster’s operation can hiccup because of any of a myriad set of reasons from bugs in HBase itself through misconfigurations — misconfiguration of HBase but also operating system misconfigurations — through to hardware problems whether it be a bug in your network card drivers or an underprovisioned RAM bus (to mention two recent examples of hardware issues that manifested as "HBase is slow"). You will also need to do a recalibration if up to this your computing has been bound to a single box. Here is one good starting point: Fallacies of Distributed Computing.

That said, you are welcome.
It’s a fun place to be.
Yours, the HBase Community.

Reporting Bugs

Please use JIRA to report non-security-related bugs.

To protect existing HBase installations from new vulnerabilities, please do not use JIRA to report security-related bugs. Instead, send your report to the mailing list private@hbase.apache.org, which allows anyone to send messages, but restricts who can read them. Someone on that list will contact you to follow up on your report.

Support and Testing Expectations

The phrases /supported/, /not supported/, /tested/, and /not tested/ occur several places throughout this guide. In the interest of clarity, here is a brief explanation of what is generally meant by these phrases, in the context of HBase.

Commercial technical support for Apache HBase is provided by many Hadoop vendors. This is not the sense in which the term /support/ is used in the context of the Apache HBase project. The Apache HBase team assumes no responsibility for your HBase clusters, your configuration, or your data.

In the context of Apache HBase, /supported/ means that HBase is designed to work in the way described, and deviation from the defined behavior or functionality should be reported as a bug.

Not Supported

In the context of Apache HBase, /not supported/ means that a use case or use pattern is not expected to work and should be considered an antipattern. If you think this designation should be reconsidered for a given feature or use pattern, file a JIRA or start a discussion on one of the mailing lists.


In the context of Apache HBase, /tested/ means that a feature is covered by unit or integration tests, and has been proven to work as expected.

Not Tested

In the context of Apache HBase, /not tested/ means that a feature or use pattern may or may notwork in a given way, and may or may not corrupt your data or cause operational issues. It is an unknown, and there are no guarantees. If you can provide proof that a feature designated as /not tested/ does work in a given way, please submit the tests and/or the metrics so that other users can gain certainty about such features or use patterns.

Getting Started

1. Introduction

Quickstart will get you up and running on a single-node, standalone instance of HBase, followed by a pseudo-distributed single-machine instance, and finally a fully-distributed cluster.

2. Quick Start - Standalone HBase

This guide describes the setup of a standalone HBase instance running against the local filesystem. This is not an appropriate configuration for a production instance of HBase, but will allow you to experiment with HBase. This section shows you how to create a table in HBase using the hbase shell CLI, insert rows into the table, perform put and scan operations against the table, enable or disable the table, and start and stop HBase. Apart from downloading HBase, this procedure should take less than 10 minutes.

Local Filesystem and Durability
The following is fixed in HBase 0.98.3 and beyond. See HBASE-11272 and HBASE-11218.

Using HBase with a local filesystem does not guarantee durability. The HDFS local filesystem implementation will lose edits if files are not properly closed. This is very likely to happen when you are experimenting with new software, starting and stopping the daemons often and not always cleanly. You need to run HBase on HDFS to ensure all writes are preserved. Running against the local filesystem is intended as a shortcut to get you familiar with how the general system works, as the very first phase of evaluation. See HBASE-3696 and its associated issues for more details about the issues of running on the local filesystem.

Loopback IP - HBase 0.94.x and earlier
The below advice is for hbase-0.94.x and older versions only. This is fixed in hbase-0.96.0 and beyond.

Prior to HBase 0.94.x, HBase expected the loopback IP address to be Ubuntu and some other distributions default to and this will cause problems for you. See Why does HBase care about /etc/hosts? for detail

Example 1. Example /etc/hosts File for Ubuntu

The following /etc/hosts file works correctly for HBase 0.94.x and earlier, on Ubuntu. Use this as a template if you run into trouble. localhost ubuntu.ubuntu-domain ubuntu

2.1. JDK Version Requirements

HBase requires that a JDK be installed. See Java for information about supported JDK versions.

2.2. Get Started with HBase

Procedure: Download, Configure, and Start HBase
  1. Choose a download site from this list of Apache Download Mirrors. Click on the suggested top link. This will take you to a mirror of HBase Releases. Click on the folder named stable and then download the binary file that ends in .tar.gz to your local filesystem. Prior to 1.x version, be sure to choose the version that corresponds with the version of Hadoop you are likely to use later (in most cases, you should choose the file for Hadoop 2, which will be called something like hbase-0.98.13-hadoop2-bin.tar.gz). Do not download the file ending in src.tar.gz for now.

  2. Extract the downloaded file, and change to the newly-created directory.

    $ tar xzvf hbase-<?eval ${project.version}?>-bin.tar.gz
    $ cd hbase-<?eval ${project.version}?>/
  3. For HBase 0.98.5 and later, you are required to set the JAVA_HOME environment variable before starting HBase. Prior to 0.98.5, HBase attempted to detect the location of Java if the variables was not set. You can set the variable via your operating system’s usual mechanism, but HBase provides a central mechanism, conf/hbase-env.sh. Edit this file, uncomment the line starting with JAVA_HOME, and set it to the appropriate location for your operating system. The JAVA_HOME variable should be set to a directory which contains the executable file bin/java. Most modern Linux operating systems provide a mechanism, such as /usr/bin/alternatives on RHEL or CentOS, for transparently switching between versions of executables such as Java. In this case, you can set JAVA_HOME to the directory containing the symbolic link to bin/java, which is usually /usr.

    These instructions assume that each node of your cluster uses the same configuration. If this is not the case, you may need to set JAVA_HOME separately for each node.
  4. Edit conf/hbase-site.xml, which is the main HBase configuration file. At this time, you only need to specify the directory on the local filesystem where HBase and ZooKeeper write data. By default, a new directory is created under /tmp. Many servers are configured to delete the contents of /tmp upon reboot, so you should store the data elsewhere. The following configuration will store HBase’s data in the hbase directory, in the home directory of the user called testuser. Paste the <property> tags beneath the <configuration> tags, which should be empty in a new HBase install.

    Example 2. Example hbase-site.xml for Standalone HBase

    You do not need to create the HBase data directory. HBase will do this for you. If you create the directory, HBase will attempt to do a migration, which is not what you want.

  5. The bin/start-hbase.sh script is provided as a convenient way to start HBase. Issue the command, and if all goes well, a message is logged to standard output showing that HBase started successfully. You can use the jps command to verify that you have one running process called HMaster. In standalone mode HBase runs all daemons within this single JVM, i.e. the HMaster, a single HRegionServer, and the ZooKeeper daemon.

    Java needs to be installed and available. If you get an error indicating that Java is not installed, but it is on your system, perhaps in a non-standard location, edit the conf/hbase-env.sh file and modify the JAVA_HOME setting to point to the directory that contains bin/java your system.
Procedure: Use HBase For the First Time
  1. Connect to HBase.

    Connect to your running instance of HBase using the hbase shell command, located in the bin/ directory of your HBase install. In this example, some usage and version information that is printed when you start HBase Shell has been omitted. The HBase Shell prompt ends with a > character.

    $ ./bin/hbase shell
  2. Display HBase Shell Help Text.

    Type help and press Enter, to display some basic usage information for HBase Shell, as well as several example commands. Notice that table names, rows, columns all must be enclosed in quote characters.

  3. Create a table.

    Use the create command to create a new table. You must specify the table name and the ColumnFamily name.

    hbase(main):001:0> create 'test', 'cf'
    0 row(s) in 0.4170 seconds
    => Hbase::Table - test
  4. List Information About your Table

    Use the list command to

    hbase(main):002:0> list 'test'
    1 row(s) in 0.0180 seconds
    => ["test"]
  5. Put data into your table.

    To put data into your table, use the put command.

    hbase(main):003:0> put 'test', 'row1', 'cf:a', 'value1'
    0 row(s) in 0.0850 seconds
    hbase(main):004:0> put 'test', 'row2', 'cf:b', 'value2'
    0 row(s) in 0.0110 seconds
    hbase(main):005:0> put 'test', 'row3', 'cf:c', 'value3'
    0 row(s) in 0.0100 seconds

    Here, we insert three values, one at a time. The first insert is at row1, column cf:a, with a value of value1. Columns in HBase are comprised of a column family prefix, cf in this example, followed by a colon and then a column qualifier suffix, a in this case.

  6. Scan the table for all data at once.

    One of the ways to get data from HBase is to scan. Use the scan command to scan the table for data. You can limit your scan, but for now, all data is fetched.

    hbase(main):006:0> scan 'test'
    ROW                                      COLUMN+CELL
     row1                                    column=cf:a, timestamp=1421762485768, value=value1
     row2                                    column=cf:b, timestamp=1421762491785, value=value2
     row3                                    column=cf:c, timestamp=1421762496210, value=value3
    3 row(s) in 0.0230 seconds
  7. Get a single row of data.

    To get a single row of data at a time, use the get command.

    hbase(main):007:0> get 'test', 'row1'
    COLUMN                                   CELL
     cf:a                                    timestamp=1421762485768, value=value1
    1 row(s) in 0.0350 seconds
  8. Disable a table.

    If you want to delete a table or change its settings, as well as in some other situations, you need to disable the table first, using the disable command. You can re-enable it using the enable command.

    hbase(main):008:0> disable 'test'
    0 row(s) in 1.1820 seconds
    hbase(main):009:0> enable 'test'
    0 row(s) in 0.1770 seconds

    Disable the table again if you tested the enable command above:

    hbase(main):010:0> disable 'test'
    0 row(s) in 1.1820 seconds
  9. Drop the table.

    To drop (delete) a table, use the drop command.

    hbase(main):011:0> drop 'test'
    0 row(s) in 0.1370 seconds
  10. Exit the HBase Shell.

    To exit the HBase Shell and disconnect from your cluster, use the quit command. HBase is still running in the background.

Procedure: Stop HBase
  1. In the same way that the bin/start-hbase.sh script is provided to conveniently start all HBase daemons, the bin/stop-hbase.sh script stops them.

    $ ./bin/stop-hbase.sh
    stopping hbase....................
  2. After issuing the command, it can take several minutes for the processes to shut down. Use the jps to be sure that the HMaster and HRegionServer processes are shut down.

2.3. Intermediate - Pseudo-Distributed Local Install

After working your way through quickstart, you can re-configure HBase to run in pseudo-distributed mode. Pseudo-distributed mode means that HBase still runs completely on a single host, but each HBase daemon (HMaster, HRegionServer, and ZooKeeper) runs as a separate process. By default, unless you configure the hbase.rootdir property as described in quickstart, your data is still stored in /tmp/. In this walk-through, we store your data in HDFS instead, assuming you have HDFS available. You can skip the HDFS configuration to continue storing your data in the local filesystem.

Hadoop Configuration

This procedure assumes that you have configured Hadoop and HDFS on your local system and/or a remote system, and that they are running and available. It also assumes you are using Hadoop 2. The guide on Setting up a Single Node Cluster in the Hadoop documentation is a good starting point.

  1. Stop HBase if it is running.

    If you have just finished quickstart and HBase is still running, stop it. This procedure will create a totally new directory where HBase will store its data, so any databases you created before will be lost.

  2. Configure HBase.

    Edit the hbase-site.xml configuration. First, add the following property. which directs HBase to run in distributed mode, with one JVM instance per daemon.


    Next, change the hbase.rootdir from the local filesystem to the address of your HDFS instance, using the hdfs://// URI syntax. In this example, HDFS is running on the localhost at port 8020.


    You do not need to create the directory in HDFS. HBase will do this for you. If you create the directory, HBase will attempt to do a migration, which is not what you want.

  3. Start HBase.

    Use the bin/start-hbase.sh command to start HBase. If your system is configured correctly, the jps command should show the HMaster and HRegionServer processes running.

  4. Check the HBase directory in HDFS.

    If everything worked correctly, HBase created its directory in HDFS. In the configuration above, it is stored in /hbase/ on HDFS. You can use the hadoop fs command in Hadoop’s bin/ directory to list this directory.

    $ ./bin/hadoop fs -ls /hbase
    Found 7 items
    drwxr-xr-x   - hbase users          0 2014-06-25 18:58 /hbase/.tmp
    drwxr-xr-x   - hbase users          0 2014-06-25 21:49 /hbase/WALs
    drwxr-xr-x   - hbase users          0 2014-06-25 18:48 /hbase/corrupt
    drwxr-xr-x   - hbase users          0 2014-06-25 18:58 /hbase/data
    -rw-r--r--   3 hbase users         42 2014-06-25 18:41 /hbase/hbase.id
    -rw-r--r--   3 hbase users          7 2014-06-25 18:41 /hbase/hbase.version
    drwxr-xr-x   - hbase users          0 2014-06-25 21:49 /hbase/oldWALs
  5. Create a table and populate it with data.

    You can use the HBase Shell to create a table, populate it with data, scan and get values from it, using the same procedure as in shell exercises.

  6. Start and stop a backup HBase Master (HMaster) server.

    Running multiple HMaster instances on the same hardware does not make sense in a production environment, in the same way that running a pseudo-distributed cluster does not make sense for production. This step is offered for testing and learning purposes only.

    The HMaster server controls the HBase cluster. You can start up to 9 backup HMaster servers, which makes 10 total HMasters, counting the primary. To start a backup HMaster, use the local-master-backup.sh. For each backup master you want to start, add a parameter representing the port offset for that master. Each HMaster uses three ports (16010, 16020, and 16030 by default). The port offset is added to these ports, so using an offset of 2, the backup HMaster would use ports 16012, 16022, and 16032. The following command starts 3 backup servers using ports 16012/16022/16032, 16013/16023/16033, and 16015/16025/16035.

    $ ./bin/local-master-backup.sh 2 3 5

    To kill a backup master without killing the entire cluster, you need to find its process ID (PID). The PID is stored in a file with a name like /tmp/hbase-USER-X-master.pid. The only contents of the file is the PID. You can use the kill -9 command to kill that PID. The following command will kill the master with port offset 1, but leave the cluster running:

    $ cat /tmp/hbase-testuser-1-master.pid |xargs kill -9
  7. Start and stop additional RegionServers

    The HRegionServer manages the data in its StoreFiles as directed by the HMaster. Generally, one HRegionServer runs per node in the cluster. Running multiple HRegionServers on the same system can be useful for testing in pseudo-distributed mode. The local-regionservers.sh command allows you to run multiple RegionServers. It works in a similar way to the local-master-backup.sh command, in that each parameter you provide represents the port offset for an instance. Each RegionServer requires two ports, and the default ports are 16020 and 16030. However, the base ports for additional RegionServers are not the default ports since the default ports are used by the HMaster, which is also a RegionServer since HBase version 1.0.0. The base ports are 16200 and 16300 instead. You can run 99 additional RegionServers that are not a HMaster or backup HMaster, on a server. The following command starts four additional RegionServers, running on sequential ports starting at 16202/16302 (base ports 16200/16300 plus 2).

    $ .bin/local-regionservers.sh start 2 3 4 5

    To stop a RegionServer manually, use the local-regionservers.sh command with the stop parameter and the offset of the server to stop.

    $ .bin/local-regionservers.sh stop 3
  8. Stop HBase.

    You can stop HBase the same way as in the quickstart procedure, using the bin/stop-hbase.sh command.

2.4. Advanced - Fully Distributed

In reality, you need a fully-distributed configuration to fully test HBase and to use it in real-world scenarios. In a distributed configuration, the cluster contains multiple nodes, each of which runs one or more HBase daemon. These include primary and backup Master instances, multiple ZooKeeper nodes, and multiple RegionServer nodes.

This advanced quickstart adds two more nodes to your cluster. The architecture will be as follows:

Table 1. Distributed Cluster Demo Architecture
Node Name Master ZooKeeper RegionServer













This quickstart assumes that each node is a virtual machine and that they are all on the same network. It builds upon the previous quickstart, Intermediate - Pseudo-Distributed Local Install, assuming that the system you configured in that procedure is now node-a. Stop HBase on node-a before continuing.

Be sure that all the nodes have full access to communicate, and that no firewall rules are in place which could prevent them from talking to each other. If you see any errors like no route to host, check your firewall.
Procedure: Configure Passwordless SSH Access

node-a needs to be able to log into node-b and node-c (and to itself) in order to start the daemons. The easiest way to accomplish this is to use the same username on all hosts, and configure password-less SSH login from node-a to each of the others.

  1. On node-a, generate a key pair.

    While logged in as the user who will run HBase, generate a SSH key pair, using the following command:

    $ ssh-keygen -t rsa

    If the command succeeds, the location of the key pair is printed to standard output. The default name of the public key is id_rsa.pub.

  2. Create the directory that will hold the shared keys on the other nodes.

    On node-b and node-c, log in as the HBase user and create a .ssh/ directory in the user’s home directory, if it does not already exist. If it already exists, be aware that it may already contain other keys.

  3. Copy the public key to the other nodes.

    Securely copy the public key from node-a to each of the nodes, by using the scp or some other secure means. On each of the other nodes, create a new file called .ssh/authorized_keys if it does not already exist, and append the contents of the id_rsa.pub file to the end of it. Note that you also need to do this for node-a itself.

    $ cat id_rsa.pub >> ~/.ssh/authorized_keys
  4. Test password-less login.

    If you performed the procedure correctly, if you SSH from node-a to either of the other nodes, using the same username, you should not be prompted for a password.

  5. Since node-b will run a backup Master, repeat the procedure above, substituting node-b everywhere you see node-a. Be sure not to overwrite your existing .ssh/authorized_keys files, but concatenate the new key onto the existing file using the >> operator rather than the > operator.

Procedure: Prepare node-a

node-a will run your primary master and ZooKeeper processes, but no RegionServers. . Stop the RegionServer from starting on node-a.

  1. Edit conf/regionservers and remove the line which contains localhost. Add lines with the hostnames or IP addresses for node-b and node-c.

    Even if you did want to run a RegionServer on node-a, you should refer to it by the hostname the other servers would use to communicate with it. In this case, that would be node-a.example.com. This enables you to distribute the configuration to each node of your cluster any hostname conflicts. Save the file.

  2. Configure HBase to use node-b as a backup master.

    Create a new file in conf/ called backup-masters, and add a new line to it with the hostname for node-b. In this demonstration, the hostname is node-b.example.com.

  3. Configure ZooKeeper

    In reality, you should carefully consider your ZooKeeper configuration. You can find out more about configuring ZooKeeper in zookeeper. This configuration will direct HBase to start and manage a ZooKeeper instance on each node of the cluster.

    On node-a, edit conf/hbase-site.xml and add the following properties.

  4. Everywhere in your configuration that you have referred to node-a as localhost, change the reference to point to the hostname that the other nodes will use to refer to node-a. In these examples, the hostname is node-a.example.com.

Procedure: Prepare node-b and node-c

node-b will run a backup master server and a ZooKeeper instance.

  1. Download and unpack HBase.

    Download and unpack HBase to node-b, just as you did for the standalone and pseudo-distributed quickstarts.

  2. Copy the configuration files from node-a to node-b.and node-c.

    Each node of your cluster needs to have the same configuration information. Copy the contents of the conf/ directory to the conf/ directory on node-b and node-c.

Procedure: Start and Test Your Cluster
  1. Be sure HBase is not running on any node.

    If you forgot to stop HBase from previous testing, you will have errors. Check to see whether HBase is running on any of your nodes by using the jps command. Look for the processes HMaster, HRegionServer, and HQuorumPeer. If they exist, kill them.

  2. Start the cluster.

    On node-a, issue the start-hbase.sh command. Your output will be similar to that below.

    $ bin/start-hbase.sh
    node-c.example.com: starting zookeeper, logging to /home/hbuser/hbase-0.98.3-hadoop2/bin/../logs/hbase-hbuser-zookeeper-node-c.example.com.out
    node-a.example.com: starting zookeeper, logging to /home/hbuser/hbase-0.98.3-hadoop2/bin/../logs/hbase-hbuser-zookeeper-node-a.example.com.out
    node-b.example.com: starting zookeeper, logging to /home/hbuser/hbase-0.98.3-hadoop2/bin/../logs/hbase-hbuser-zookeeper-node-b.example.com.out
    starting master, logging to /home/hbuser/hbase-0.98.3-hadoop2/bin/../logs/hbase-hbuser-master-node-a.example.com.out
    node-c.example.com: starting regionserver, logging to /home/hbuser/hbase-0.98.3-hadoop2/bin/../logs/hbase-hbuser-regionserver-node-c.example.com.out
    node-b.example.com: starting regionserver, logging to /home/hbuser/hbase-0.98.3-hadoop2/bin/../logs/hbase-hbuser-regionserver-node-b.example.com.out
    node-b.example.com: starting master, logging to /home/hbuser/hbase-0.98.3-hadoop2/bin/../logs/hbase-hbuser-master-nodeb.example.com.out

    ZooKeeper starts first, followed by the master, then the RegionServers, and finally the backup masters.

  3. Verify that the processes are running.

    On each node of the cluster, run the jps command and verify that the correct processes are running on each server. You may see additional Java processes running on your servers as well, if they are used for other purposes.

    Example 3. node-a jps Output
    $ jps
    20355 Jps
    20071 HQuorumPeer
    20137 HMaster
    Example 4. node-b jps Output
    $ jps
    15930 HRegionServer
    16194 Jps
    15838 HQuorumPeer
    16010 HMaster
    Example 5. node-a jps Output
    $ jps
    13901 Jps
    13639 HQuorumPeer
    13737 HRegionServer
    ZooKeeper Process Name

    The HQuorumPeer process is a ZooKeeper instance which is controlled and started by HBase. If you use ZooKeeper this way, it is limited to one instance per cluster node, , and is appropriate for testing only. If ZooKeeper is run outside of HBase, the process is called QuorumPeer. For more about ZooKeeper configuration, including using an external ZooKeeper instance with HBase, see zookeeper.

  4. Browse to the Web UI.

    Web UI Port Changes
    Web UI Port Changes

    In HBase newer than 0.98.x, the HTTP ports used by the HBase Web UI changed from 60010 for the Master and 60030 for each RegionServer to 16010 for the Master and 16030 for the RegionServer.

    If everything is set up correctly, you should be able to connect to the UI for the Master http://node-a.example.com:16010/ or the secondary master at http://node-b.example.com:16010/ for the secondary master, using a web browser. If you can connect via localhost but not from another host, check your firewall rules. You can see the web UI for each of the RegionServers at port 16030 of their IP addresses, or by clicking their links in the web UI for the Master.

  5. Test what happens when nodes or services disappear.

    With a three-node cluster like you have configured, things will not be very resilient. Still, you can test what happens when the primary Master or a RegionServer disappears, by killing the processes and watching the logs.

2.5. Where to go next

The next chapter, configuration, gives more information about the different HBase run modes, system requirements for running HBase, and critical configuration areas for setting up a distributed HBase cluster.

Apache HBase Configuration

This chapter expands upon the Getting Started chapter to further explain configuration of Apache HBase. Please read this chapter carefully, especially the Basic Prerequisites to ensure that your HBase testing and deployment goes smoothly, and prevent data loss. Familiarize yourself with Support and Testing Expectations as well.

3. Configuration Files

Apache HBase uses the same configuration system as Apache Hadoop. All configuration files are located in the conf/ directory, which needs to be kept in sync for each node on your cluster.

HBase Configuration File Descriptions

Not present by default. A plain-text file which lists hosts on which the Master should start a backup Master process, one host per line.


Used to connect HBase Hadoop’s Metrics2 framework. See the Hadoop Wiki entry for more information on Metrics2. Contains only commented-out examples by default.

hbase-env.cmd and hbase-env.sh

Script for Windows and Linux / Unix environments to set up the working environment for HBase, including the location of Java, Java options, and other environment variables. The file contains many commented-out examples to provide guidance.


The default policy configuration file used by RPC servers to make authorization decisions on client requests. Only used if HBase security is enabled.


The main HBase configuration file. This file specifies configuration options which override HBase’s default configuration. You can view (but do not edit) the default configuration file at docs/hbase-default.xml. You can also view the entire effective configuration for your cluster (defaults and overrides) in the HBase Configuration tab of the HBase Web UI.


Configuration file for HBase logging via log4j.


A plain-text file containing a list of hosts which should run a RegionServer in your HBase cluster. By default this file contains the single entry localhost. It should contain a list of hostnames or IP addresses, one per line, and should only contain localhost if each node in your cluster will run a RegionServer on its localhost interface.

Checking XML Validity

When you edit XML, it is a good idea to use an XML-aware editor to be sure that your syntax is correct and your XML is well-formed. You can also use the xmllint utility to check that your XML is well-formed. By default, xmllint re-flows and prints the XML to standard output. To check for well-formedness and only print output if errors exist, use the command xmllint -noout filename.xml.

Keep Configuration In Sync Across the Cluster

When running in distributed mode, after you make an edit to an HBase configuration, make sure you copy the content of the conf/ directory to all nodes of the cluster. HBase will not do this for you. Use rsync, scp, or another secure mechanism for copying the configuration files to your nodes. For most configuration, a restart is needed for servers to pick up changes An exception is dynamic configuration. to be described later below.

4. Basic Prerequisites

This section lists required services and some required system configuration.


The following table summarizes the recommendation of the HBase community wrt deploying on various Java versions. A symbol is meant to indicate a base level of testing and willingness to help diagnose and address issues you might run into. Similarly, an entry of or generally means that should you run into an issue the community is likely to ask you to change the Java environment before proceeding to help. In some cases, specific guidance on limitations (e.g. whether compiling / unit tests work, specific operational issues, etc) will also be noted.

Long Term Support JDKs are recommended

HBase recommends downstream users rely on JDK releases that are marked as Long Term Supported (LTS) either from the OpenJDK project or vendors. As of March 2018 that means Java 8 is the only applicable version and that the next likely version to see testing will be Java 11 near Q3 2018.

Table 2. Java support by release line
HBase Version JDK 7 JDK 8 JDK 9 (Non-LTS) JDK 10 (Non-LTS) JDK 11



In HBase 1.2 and newer, you must set JAVA_HOME on each node of your cluster. hbase-env.sh provides a handy mechanism to do this.
Operating System Utilities

HBase uses the Secure Shell (ssh) command and utilities extensively to communicate between cluster nodes. Each server in the cluster must be running ssh so that the Hadoop and HBase daemons can be managed. You must be able to connect to all nodes via SSH, including the local node, from the Master as well as any backup Master, using a shared key rather than a password. You can see the basic methodology for such a set-up in Linux or Unix systems at "Procedure: Configure Passwordless SSH Access". If your cluster nodes use OS X, see the section, SSH: Setting up Remote Desktop and Enabling Self-Login on the Hadoop wiki.


HBase uses the local hostname to self-report its IP address. Both forward and reverse DNS resolving must work in versions of HBase previous to 0.92.0. The hadoop-dns-checker tool can be used to verify DNS is working correctly on the cluster. The project README file provides detailed instructions on usage.

Loopback IP

Prior to hbase-0.96.0, HBase only used the IP address to refer to localhost, and this could not be configured. See Loopback IP for more details.


The clocks on cluster nodes should be synchronized. A small amount of variation is acceptable, but larger amounts of skew can cause erratic and unexpected behavior. Time synchronization is one of the first things to check if you see unexplained problems in your cluster. It is recommended that you run a Network Time Protocol (NTP) service, or another time-synchronization mechanism, on your cluster, and that all nodes look to the same service for time synchronization. See the Basic NTP Configuration at The Linux Documentation Project (TLDP) to set up NTP.

Limits on Number of Files and Processes (ulimit)

Apache HBase is a database. It requires the ability to open a large number of files at once. Many Linux distributions limit the number of files a single user is allowed to open to 1024 (or 256 on older versions of OS X). You can check this limit on your servers by running the command ulimit -n when logged in as the user which runs HBase. See the Troubleshooting section for some of the problems you may experience if the limit is too low. You may also notice errors such as the following:

2010-04-06 03:04:37,542 INFO org.apache.hadoop.hdfs.DFSClient: Exception increateBlockOutputStream java.io.EOFException
2010-04-06 03:04:37,542 INFO org.apache.hadoop.hdfs.DFSClient: Abandoning block blk_-6935524980745310745_1391901

It is recommended to raise the ulimit to at least 10,000, but more likely 10,240, because the value is usually expressed in multiples of 1024. Each ColumnFamily has at least one StoreFile, and possibly more than six StoreFiles if the region is under load. The number of open files required depends upon the number of ColumnFamilies and the number of regions. The following is a rough formula for calculating the potential number of open files on a RegionServer.

Calculate the Potential Number of Open Files
(StoreFiles per ColumnFamily) x (regions per RegionServer)

For example, assuming that a schema had 3 ColumnFamilies per region with an average of 3 StoreFiles per ColumnFamily, and there are 100 regions per RegionServer, the JVM will open 3 * 3 * 100 = 900 file descriptors, not counting open JAR files, configuration files, and others. Opening a file does not take many resources, and the risk of allowing a user to open too many files is minimal.

Another related setting is the number of processes a user is allowed to run at once. In Linux and Unix, the number of processes is set using the ulimit -u command. This should not be confused with the nproc command, which controls the number of CPUs available to a given user. Under load, a ulimit -u that is too low can cause OutOfMemoryError exceptions. See Jack Levin’s major HDFS issues thread on the hbase-users mailing list, from 2011.

Configuring the maximum number of file descriptors and processes for the user who is running the HBase process is an operating system configuration, rather than an HBase configuration. It is also important to be sure that the settings are changed for the user that actually runs HBase. To see which user started HBase, and that user’s ulimit configuration, look at the first line of the HBase log for that instance. A useful read setting config on your hadoop cluster is Aaron Kimball’s Configuration Parameters: What can you just ignore?

Example 6. ulimit Settings on Ubuntu

To configure ulimit settings on Ubuntu, edit /etc/security/limits.conf, which is a space-delimited file with four columns. Refer to the man page for limits.conf for details about the format of this file. In the following example, the first line sets both soft and hard limits for the number of open files (nofile) to 32768 for the operating system user with the username hadoop. The second line sets the number of processes to 32000 for the same user.

hadoop  -       nofile  32768
hadoop  -       nproc   32000

The settings are only applied if the Pluggable Authentication Module (PAM) environment is directed to use them. To configure PAM to use these limits, be sure that the /etc/pam.d/common-session file contains the following line:

session required  pam_limits.so
Linux Shell

All of the shell scripts that come with HBase rely on the GNU Bash shell.


Prior to HBase 0.96, testing for running HBase on Microsoft Windows was limited. Running a on Windows nodes is not recommended for production systems.

4.1. Hadoop

The following table summarizes the versions of Hadoop supported with each version of HBase. Based on the version of HBase, you should select the most appropriate version of Hadoop. You can use Apache Hadoop, or a vendor’s distribution of Hadoop. No distinction is made here. See the Hadoop wiki for information about vendors of Hadoop.

Hadoop 2.x is recommended.

Hadoop 2.x is faster and includes features, such as short-circuit reads, which will help improve your HBase random read profile. Hadoop 2.x also includes important bug fixes that will improve your overall HBase experience. HBase 0.98 drops support for Hadoop 1.0, deprecates use of Hadoop 1.1+, and HBase 1.0 will not support Hadoop 1.x.

Use the following legend to interpret this table:

Hadoop version support matrix
  • = Tested to be fully-functional

  • = Known to not be fully-functional

  • = Not tested, may/may-not function

HBase-1.2.x, HBase-1.3.x HBase-1.4.x HBase-2.0.x HBase-2.1.x
















Hadoop 2.6.x

Hadoop distributions based on the 2.6.x line must have HADOOP-11710 applied if you plan to run HBase on top of an HDFS Encryption Zone. Failure to do so will result in cluster failure and data loss. This patch is present in Apache Hadoop releases 2.6.1+.

Hadoop 2.7.x

Hadoop version 2.7.0 is not tested or supported as the Hadoop PMC has explicitly labeled that release as not being stable.

Replace the Hadoop Bundled With HBase!

Because HBase depends on Hadoop, it bundles an instance of the Hadoop jar under its lib directory. The bundled jar is ONLY for use in standalone mode. In distributed mode, it is critical that the version of Hadoop that is out on your cluster match what is under HBase. Replace the hadoop jar found in the HBase lib directory with the hadoop jar you are running on your cluster to avoid version mismatch issues. Make sure you replace the jar in HBase everywhere on your cluster. Hadoop version mismatch issues have various manifestations but often all looks like its hung up.

4.1.1. Apache HBase 0.94 with Hadoop 2

To get 0.94.x to run on Hadoop 2.2.0, you need to change the hadoop 2 and protobuf versions in the pom.xml: Here is a diff with pom.xml changes:

$ svn diff pom.xml
Index: pom.xml
--- pom.xml     (revision 1545157)
+++ pom.xml     (working copy)
@@ -1034,7 +1034,7 @@
-    <protobuf.version>2.4.0a</protobuf.version>
+    <protobuf.version>2.5.0</protobuf.version>
@@ -2241,7 +2241,7 @@
-        <hadoop.version>2.0.0-alpha</hadoop.version>
+        <hadoop.version>2.2.0</hadoop.version>

The next step is to regenerate Protobuf files and assuming that the Protobuf has been installed:

  • Go to the HBase root folder, using the command line;

  • Type the following commands:

    $ protoc -Isrc/main/protobuf --java_out=src/main/java src/main/protobuf/hbase.proto
    $ protoc -Isrc/main/protobuf --java_out=src/main/java src/main/protobuf/ErrorHandling.proto

Building against the hadoop 2 profile by running something like the following command:

$  mvn clean install assembly:single -Dhadoop.profile=2.0 -DskipTests

4.1.2. Apache HBase 0.92 and 0.94

HBase 0.92 and 0.94 versions can work with Hadoop versions, 0.20.205, 0.22.x, 1.0.x, and 1.1.x. HBase-0.94 can additionally work with Hadoop-0.23.x and 2.x, but you may have to recompile the code using the specific maven profile (see top level pom.xml)

4.1.3. Apache HBase 0.96

As of Apache HBase 0.96.x, Apache Hadoop 1.0.x at least is required. Hadoop 2 is strongly encouraged (faster but also has fixes that help MTTR). We will no longer run properly on older Hadoops such as 0.20.205 or branch-0.20-append. Do not move to Apache HBase 0.96.x if you cannot upgrade your Hadoop. See HBase, mail # dev - DISCUSS: Have hbase require at least hadoop 1.0.0 in hbase 0.96.0?

4.1.4. Hadoop versions 0.20.x - 1.x

HBase will lose data unless it is running on an HDFS that has a durable sync implementation. DO NOT use Hadoop 0.20.2, Hadoop, and Hadoop which DO NOT have this attribute. Currently only Hadoop versions 0.20.205.x or any release in excess of this version — this includes hadoop-1.0.0 — have a working, durable sync. The Cloudera blog post An update on Apache Hadoop 1.0 by Charles Zedlweski has a nice exposition on how all the Hadoop versions relate. It’s worth checking out if you are having trouble making sense of the Hadoop version morass.

Sync has to be explicitly enabled by setting dfs.support.append equal to true on both the client side — in hbase-site.xml — and on the serverside in hdfs-site.xml (The sync facility HBase needs is a subset of the append code path).


You will have to restart your cluster after making this edit. Ignore the chicken-little comment you’ll find in the hdfs-default.xml in the description for the dfs.support.append configuration.

4.1.5. Apache HBase on Secure Hadoop

Apache HBase will run on any Hadoop 0.20.x that incorporates Hadoop security features as long as you do as suggested above and replace the Hadoop jar that ships with HBase with the secure version. If you want to read more about how to setup Secure HBase, see hbase.secure.configuration.

4.1.6. dfs.datanode.max.transfer.threads

An HDFS DataNode has an upper bound on the number of files that it will serve at any one time. Before doing any loading, make sure you have configured Hadoop’s conf/hdfs-site.xml, setting the dfs.datanode.max.transfer.threads value to at least the following:


Be sure to restart your HDFS after making the above configuration.

Not having this configuration in place makes for strange-looking failures. One manifestation is a complaint about missing blocks. For example:

10/12/08 20:10:31 INFO hdfs.DFSClient: Could not obtain block
          blk_XXXXXXXXXXXXXXXXXXXXXX_YYYYYYYY from any node: java.io.IOException: No live nodes
          contain current block. Will get new block locations from namenode and retry...

See also casestudies.max.transfer.threads and note that this property was previously known as dfs.datanode.max.xcievers (e.g. Hadoop HDFS: Deceived by Xciever).

4.2. ZooKeeper Requirements

ZooKeeper 3.4.x is required as of HBase 1.0.0. HBase makes use of the multi functionality that is only available since 3.4.0 (The useMulti configuration option defaults to true in HBase 1.0.0). See HBASE-12241 (The crash of regionServer when taking deadserver’s replication queue breaks replication) and HBASE-6775 (Use ZK.multi when available for HBASE-6710 0.92/0.94 compatibility fix) for background.

5. HBase run modes: Standalone and Distributed

HBase has two run modes: standalone and distributed. Out of the box, HBase runs in standalone mode. Whatever your mode, you will need to configure HBase by editing files in the HBase conf directory. At a minimum, you must edit conf/hbase-env.sh to tell HBase which java to use. In this file you set HBase environment variables such as the heapsize and other options for the JVM, the preferred location for log files, etc. Set JAVA_HOME to point at the root of your java install.

5.1. Standalone HBase

This is the default mode. Standalone mode is what is described in the quickstart section. In standalone mode, HBase does not use HDFS — it uses the local filesystem instead — and it runs all HBase daemons and a local ZooKeeper all up in the same JVM. ZooKeeper binds to a well known port so clients may talk to HBase.

5.2. Distributed

Distributed mode can be subdivided into distributed but all daemons run on a single node — a.k.a. pseudo-distributed — and fully-distributed where the daemons are spread across all nodes in the cluster. The pseudo-distributed vs. fully-distributed nomenclature comes from Hadoop.

Pseudo-distributed mode can run against the local filesystem or it can run against an instance of the Hadoop Distributed File System (HDFS). Fully-distributed mode can ONLY run on HDFS. See the Hadoop documentation for how to set up HDFS. A good walk-through for setting up HDFS on Hadoop 2 can be found at http://www.alexjf.net/blog/distributed-systems/hadoop-yarn-installation-definitive-guide.

5.2.1. Pseudo-distributed

Pseudo-Distributed Quickstart

A quickstart has been added to the quickstart chapter. See quickstart-pseudo. Some of the information that was originally in this section has been moved there.

A pseudo-distributed mode is simply a fully-distributed mode run on a single host. Use this configuration testing and prototyping on HBase. Do not use this configuration for production nor for evaluating HBase performance.

5.3. Fully-distributed

By default, HBase runs in standalone mode. Both standalone mode and pseudo-distributed mode are provided for the purposes of small-scale testing. For a production environment, distributed mode is appropriate. In distributed mode, multiple instances of HBase daemons run on multiple servers in the cluster.

Just as in pseudo-distributed mode, a fully distributed configuration requires that you set the hbase-cluster.distributed property to true. Typically, the hbase.rootdir is configured to point to a highly-available HDFS filesystem.

In addition, the cluster is configured so that multiple cluster nodes enlist as RegionServers, ZooKeeper QuorumPeers, and backup HMaster servers. These configuration basics are all demonstrated in quickstart-fully-distributed.

Distributed RegionServers

Typically, your cluster will contain multiple RegionServers all running on different servers, as well as primary and backup Master and ZooKeeper daemons. The conf/regionservers file on the master server contains a list of hosts whose RegionServers are associated with this cluster. Each host is on a separate line. All hosts listed in this file will have their RegionServer processes started and stopped when the master server starts or stops.

ZooKeeper and HBase

See the ZooKeeper section for ZooKeeper setup instructions for HBase.

Example 7. Example Distributed HBase Cluster

This is a bare-bones conf/hbase-site.xml for a distributed HBase cluster. A cluster that is used for real-world work would contain more custom configuration parameters. Most HBase configuration directives have default values, which are used unless the value is overridden in the hbase-site.xml. See "Configuration Files" for more information.


This is an example conf/regionservers file, which contains a list of nodes that should run a RegionServer in the cluster. These nodes need HBase installed and they need to use the same contents of the conf/ directory as the Master server


This is an example conf/backup-masters file, which contains a list of each node that should run a backup Master instance. The backup Master instances will sit idle unless the main Master becomes unavailable.

Distributed HBase Quickstart

See quickstart-fully-distributed for a walk-through of a simple three-node cluster configuration with multiple ZooKeeper, backup HMaster, and RegionServer instances.

Procedure: HDFS Client Configuration
  1. Of note, if you have made HDFS client configuration changes on your Hadoop cluster, such as configuration directives for HDFS clients, as opposed to server-side configurations, you must use one of the following methods to enable HBase to see and use these configuration changes:

    1. Add a pointer to your HADOOP_CONF_DIR to the HBASE_CLASSPATH environment variable in hbase-env.sh.

    2. Add a copy of hdfs-site.xml (or hadoop-site.xml) or, better, symlinks, under ${HBASE_HOME}/conf, or

    3. if only a small set of HDFS client configurations, add them to hbase-site.xml.

An example of such an HDFS client configuration is dfs.replication. If for example, you want to run with a replication factor of 5, HBase will create files with the default of 3 unless you do the above to make the configuration available to HBase.

6. Running and Confirming Your Installation

Make sure HDFS is running first. Start and stop the Hadoop HDFS daemons by running bin/start-hdfs.sh over in the HADOOP_HOME directory. You can ensure it started properly by testing the put and get of files into the Hadoop filesystem. HBase does not normally use the MapReduce or YARN daemons. These do not need to be started.

If you are managing your own ZooKeeper, start it and confirm it’s running, else HBase will start up ZooKeeper for you as part of its start process.

Start HBase with the following command:


Run the above from the HBASE_HOME directory.

You should now have a running HBase instance. HBase logs can be found in the logs subdirectory. Check them out especially if HBase had trouble starting.

HBase also puts up a UI listing vital attributes. By default it’s deployed on the Master host at port 16010 (HBase RegionServers listen on port 16020 by default and put up an informational HTTP server at port 16030). If the Master is running on a host named master.example.org on the default port, point your browser at http://master.example.org:16010 to see the web interface.

Prior to HBase 0.98 the master UI was deployed on port 60010, and the HBase RegionServers UI on port 60030.

Once HBase has started, see the shell exercises section for how to create tables, add data, scan your insertions, and finally disable and drop your tables.

To stop HBase after exiting the HBase shell enter

$ ./bin/stop-hbase.sh
stopping hbase...............

Shutdown can take a moment to complete. It can take longer if your cluster is comprised of many machines. If you are running a distributed operation, be sure to wait until HBase has shut down completely before stopping the Hadoop daemons.

7. Default Configuration

7.1. hbase-site.xml and hbase-default.xml

Just as in Hadoop where you add site-specific HDFS configuration to the hdfs-site.xml file, for HBase, site specific customizations go into the file conf/hbase-site.xml. For the list of configurable properties, see hbase default configurations below or view the raw hbase-default.xml source file in the HBase source code at src/main/resources.

Not all configuration options make it out to hbase-default.xml. Configuration that it is thought rare anyone would change can exist only in code; the only way to turn up such configurations is via a reading of the source code itself.

Currently, changes here will require a cluster restart for HBase to notice the change.

7.2. HBase Default Configuration

The documentation below is generated using the default hbase configuration file, hbase-default.xml, as source.


Temporary directory on the local filesystem. Change this setting to point to a location more permanent than '/tmp', the usual resolve for java.io.tmpdir, as the '/tmp' directory is cleared on machine restart.




The directory shared by region servers and into which HBase persists. The URL should be 'fully-qualified' to include the filesystem scheme. For example, to specify the HDFS directory '/hbase' where the HDFS instance’s namenode is running at namenode.example.org on port 9000, set this value to: hdfs://namenode.example.org:9000/hbase. By default, we write to whatever ${hbase.tmp.dir} is set too — usually /tmp — so change this configuration or else all data will be lost on machine restart.




A staging directory in default file system (HDFS) for keeping temporary data.




A staging directory in default file system (HDFS) for bulk loading.




The mode the cluster will be in. Possible values are false for standalone mode and true for distributed mode. If false, startup will run all HBase and ZooKeeper daemons together in the one JVM.




Comma separated list of servers in the ZooKeeper ensemble (This config. should have been named hbase.zookeeper.ensemble). For example, "host1.mydomain.com,host2.mydomain.com,host3.mydomain.com". By default this is set to localhost for local and pseudo-distributed modes of operation. For a fully-distributed setup, this should be set to a full list of ZooKeeper ensemble servers. If HBASE_MANAGES_ZK is set in hbase-env.sh this is the list of servers which hbase will start/stop ZooKeeper on as part of cluster start/stop. Client-side, we will take this list of ensemble members and put it together with the hbase.zookeeper.clientPort config. and pass it into zookeeper constructor as the connectString parameter.




Directory on the local filesystem to be used as a local storage.




The port the HBase Master should bind to.




The port for the HBase Master web UI. Set to -1 if you do not want a UI instance run.




The bind address for the HBase Master web UI



A comma-separated list of BaseLogCleanerDelegate invoked by the LogsCleaner service. These WAL cleaners are called in order, so put the cleaner that prunes the most files in front. To implement your own BaseLogCleanerDelegate, just put it in HBase’s classpath and add the fully qualified class name here. Always add the above default log cleaners in the list.




Maximum time a WAL can stay in the .oldlogdir directory, after which it will be cleaned by a Master thread.




A comma-separated list of BaseHFileCleanerDelegate invoked by the HFileCleaner service. These HFiles cleaners are called in order, so put the cleaner that prunes the most files in front. To implement your own BaseHFileCleanerDelegate, just put it in HBase’s classpath and add the fully qualified class name here. Always add the above default log cleaners in the list as they will be overwritten in hbase-site.xml.




Timeout value for the Catalog Janitor from the master to META.




Whether or not the Master listens to the Master web UI port (hbase.master.info.port) and redirects requests to the web UI server shared by the Master and RegionServer.




The port the HBase RegionServer binds to.




The port for the HBase RegionServer web UI Set to -1 if you do not want the RegionServer UI to run.




The address for the HBase RegionServer web UI



Whether or not the Master or RegionServer UI should search for a port to bind to. Enables automatic port search if hbase.regionserver.info.port is already in use. Useful for testing, turned off by default.




Count of RPC Listener instances spun up on RegionServers. Same property is used by the Master for count of master handlers.




Factor to determine the number of call queues. A value of 0 means a single queue shared between all the handlers. A value of 1 means that each handler has its own queue.




Split the call queues into read and write queues. The specified interval (which should be between 0.0 and 1.0) will be multiplied by the number of call queues. A value of 0 indicate to not split the call queues, meaning that both read and write requests will be pushed to the same set of queues. A value lower than 0.5 means that there will be less read queues than write queues. A value of 0.5 means there will be the same number of read and write queues. A value greater than 0.5 means that there will be more read queues than write queues. A value of 1.0 means that all the queues except one are used to dispatch read requests. Example: Given the total number of call queues being 10 a read.ratio of 0 means that: the 10 queues will contain both read/write requests. a read.ratio of 0.3 means that: 3 queues will contain only read requests and 7 queues will contain only write requests. a read.ratio of 0.5 means that: 5 queues will contain only read requests and 5 queues will contain only write requests. a read.ratio of 0.8 means that: 8 queues will contain only read requests and 2 queues will contain only write requests. a read.ratio of 1 means that: 9 queues will contain only read requests and 1 queues will contain only write requests.




Given the number of read call queues, calculated from the total number of call queues multiplied by the callqueue.read.ratio, the scan.ratio property will split the read call queues into small-read and long-read queues. A value lower than 0.5 means that there will be less long-read queues than short-read queues. A value of 0.5 means that there will be the same number of short-read and long-read queues. A value greater than 0.5 means that there will be more long-read queues than short-read queues A value of 0 or 1 indicate to use the same set of queues for gets and scans. Example: Given the total number of read call queues being 8 a scan.ratio of 0 or 1 means that: 8 queues will contain both long and short read requests. a scan.ratio of 0.3 means that: 2 queues will contain only long-read requests and 6 queues will contain only short-read requests. a scan.ratio of 0.5 means that: 4 queues will contain only long-read requests and 4 queues will contain only short-read requests. a scan.ratio of 0.8 means that: 6 queues will contain only long-read requests and 2 queues will contain only short-read requests.




Interval between messages from the RegionServer to Master in milliseconds.




Period at which we will roll the commit log regardless of how many edits it has.




The number of consecutive WAL close errors we will allow before triggering a server abort. A setting of 0 will cause the region server to abort if closing the current WAL writer fails during log rolling. Even a small value (2 or 3) will allow a region server to ride over transient HDFS errors.




The WAL file reader implementation.




The WAL file writer implementation.




Maximum size of all memstores in a region server before new updates are blocked and flushes are forced. Defaults to 40% of heap (0.4). Updates are blocked and flushes are forced until size of all memstores in a region server hits hbase.regionserver.global.memstore.size.lower.limit. The default value in this configuration has been intentionally left emtpy in order to honor the old hbase.regionserver.global.memstore.upperLimit property if present.




Maximum size of all memstores in a region server before flushes are forced. Defaults to 95% of hbase.regionserver.global.memstore.size (0.95). A 100% value for this value causes the minimum possible flushing to occur when updates are blocked due to memstore limiting. The default value in this configuration has been intentionally left emtpy in order to honor the old hbase.regionserver.global.memstore.lowerLimit property if present.




Maximum amount of time an edit lives in memory before being automatically flushed. Default 1 hour. Set it to 0 to disable automatic flushing.




Timeout value for the Catalog Janitor from the regionserver to META.




The name of the Network Interface from which a region server should report its IP address.




The host name or IP address of the name server (DNS) which a region server should use to determine the host name used by the master for communication and display purposes.




A split policy determines when a region should be split. The various other split policies that are available currently are ConstantSizeRegionSplitPolicy, DisabledRegionSplitPolicy, DelimitedKeyPrefixRegionSplitPolicy, KeyPrefixRegionSplitPolicy etc.




Limit for the number of regions after which no more region splitting should take place. This is not hard limit for the number of regions but acts as a guideline for the regionserver to stop splitting after a certain limit. Default is set to 1000.




ZooKeeper session timeout in milliseconds. It is used in two different ways. First, this value is used in the ZK client that HBase uses to connect to the ensemble. It is also used by HBase when it starts a ZK server and it is passed as the 'maxSessionTimeout'. See http://hadoop.apache.org/zookeeper/docs/current/zookeeperProgrammers.html#ch_zkSessions. For example, if a HBase region server connects to a ZK ensemble that’s also managed by HBase, then the session timeout will be the one specified by this configuration. But, a region server that connects to an ensemble managed with a different configuration will be subjected that ensemble’s maxSessionTimeout. So, even though HBase might propose using 90 seconds, the ensemble can have a max timeout lower than this and it will take precedence. The current default that ZK ships with is 40 seconds, which is lower than HBase’s.




Root ZNode for HBase in ZooKeeper. All of HBase’s ZooKeeper files that are configured with a relative path will go under this node. By default, all of HBase’s ZooKeeper file path are configured with a relative path, so they will all go under this directory unless changed.




Path to ZNode holding root region location. This is written by the master and read by clients and region servers. If a relative path is given, the parent folder will be ${zookeeper.znode.parent}. By default, this means the root location is stored at /hbase/root-region-server.




Root ZNode for access control lists.




The name of the Network Interface from which a ZooKeeper server should report its IP address.




The host name or IP address of the name server (DNS) which a ZooKeeper server should use to determine the host name used by the master for communication and display purposes.




Port used by ZooKeeper peers to talk to each other. See http://hadoop.apache.org/zookeeper/docs/r3.1.1/zookeeperStarted.html#sc_RunningReplicatedZooKeeper for more information.




Port used by ZooKeeper for leader election. See http://hadoop.apache.org/zookeeper/docs/r3.1.1/zookeeperStarted.html#sc_RunningReplicatedZooKeeper for more information.




Instructs HBase to make use of ZooKeeper’s multi-update functionality. This allows certain ZooKeeper operations to complete more quickly and prevents some issues with rare Replication failure scenarios (see the release note of HBASE-2611 for an example). IMPORTANT: only set this to true if all ZooKeeper servers in the cluster are on version 3.4+ and will not be downgraded. ZooKeeper versions before 3.4 do not support multi-update and will not fail gracefully if multi-update is invoked (see ZOOKEEPER-1495).




Set to true to allow HBaseConfiguration to read the zoo.cfg file for ZooKeeper properties. Switching this to true is not recommended, since the functionality of reading ZK properties from a zoo.cfg file has been deprecated.




Property from ZooKeeper’s config zoo.cfg. The number of ticks that the initial synchronization phase can take.




Property from ZooKeeper’s config zoo.cfg. The number of ticks that can pass between sending a request and getting an acknowledgment.




Property from ZooKeeper’s config zoo.cfg. The directory where the snapshot is stored.




Property from ZooKeeper’s config zoo.cfg. The port at which the clients will connect.




Property from ZooKeeper’s config zoo.cfg. Limit on number of concurrent connections (at the socket level) that a single client, identified by IP address, may make to a single member of the ZooKeeper ensemble. Set high to avoid zk connection issues running standalone and pseudo-distributed.




Default size of the HTable client write buffer in bytes. A bigger buffer takes more memory — on both the client and server side since server instantiates the passed write buffer to process it — but a larger buffer size reduces the number of RPCs made. For an estimate of server-side memory-used, evaluate hbase.client.write.buffer * hbase.regionserver.handler.count




General client pause value. Used mostly as value to wait before running a retry of a failed get, region lookup, etc. See hbase.client.retries.number for description of how we backoff from this initial pause amount and how this pause works w/ retries.




Maximum retries. Used as maximum for all retryable operations such as the getting of a cell’s value, starting a row update, etc. Retry interval is a rough function based on hbase.client.pause. At first we retry at this interval but then with backoff, we pretty quickly reach retrying every ten seconds. See HConstants#RETRY_BACKOFF for how the backup ramps up. Change this setting and hbase.client.pause to suit your workload.




The maximum number of concurrent tasks a single HTable instance will send to the cluster.




The maximum number of concurrent tasks a single HTable instance will send to a single region server.




The maximum number of concurrent connections the client will maintain to a single Region. That is, if there is already hbase.client.max.perregion.tasks writes in progress for this region, new puts won’t be sent to this region until some writes finishes.




Number of rows that we try to fetch when calling next on a scanner if it is not served from (local, client) memory. This configuration works together with hbase.client.scanner.max.result.size to try and use the network efficiently. The default value is Integer.MAX_VALUE by default so that the network will fill the chunk size defined by hbase.client.scanner.max.result.size rather than be limited by a particular number of rows since the size of rows varies table to table. If you know ahead of time that you will not require more than a certain number of rows from a scan, this configuration should be set to that row limit via Scan#setCaching. Higher caching values will enable faster scanners but will eat up more memory and some calls of next may take longer and longer times when the cache is empty. Do not set this value such that the time between invocations is greater than the scanner timeout; i.e. hbase.client.scanner.timeout.period




Specifies the combined maximum allowed size of a KeyValue instance. This is to set an upper boundary for a single entry saved in a storage file. Since they cannot be split it helps avoiding that a region cannot be split any further because the data is too large. It seems wise to set this to a fraction of the maximum region size. Setting it to zero or less disables the check.




Client scanner lease period in milliseconds.






Maximum retries. This is maximum number of iterations to atomic bulk loads are attempted in the face of splitting operations 0 means never give up.




Period at which the region balancer runs in the Master.




Period at which the region normalizer runs in the Master.




Rebalance if any regionserver has average + (average * slop) regions.




Time to sleep in between searches for work (in milliseconds). Used as sleep interval by service threads such as log roller.




How many time to retry attempting to write a version file before just aborting. Each attempt is seperated by the hbase.server.thread.wakefrequency milliseconds.




Memstore will be flushed to disk if size of the memstore exceeds this number of bytes. Value is checked by a thread that runs every hbase.server.thread.wakefrequency.




If FlushLargeStoresPolicy is used, then every time that we hit the total memstore limit, we find out all the column families whose memstores exceed this value, and only flush them, while retaining the others whose memstores are lower than this limit. If none of the families have their memstore size more than this, all the memstores will be flushed (just as usual). This value should be less than half of the total memstore threshold (hbase.hregion.memstore.flush.size).




If the memstores in a region are this size or larger when we go to close, run a "pre-flush" to clear out memstores before we put up the region closed flag and take the region offline. On close, a flush is run under the close flag to empty memory. During this time the region is offline and we are not taking on any writes. If the memstore content is large, this flush could take a long time to complete. The preflush is meant to clean out the bulk of the memstore before putting up the close flag and taking the region offline so the flush that runs under the close flag has little to do.




Block updates if memstore has hbase.hregion.memstore.block.multiplier times hbase.hregion.memstore.flush.size bytes. Useful preventing runaway memstore during spikes in update traffic. Without an upper-bound, memstore fills such that when it flushes the resultant flush files take a long time to compact or split, or worse, we OOME.




Enables the MemStore-Local Allocation Buffer, a feature which works to prevent heap fragmentation under heavy write loads. This can reduce the frequency of stop-the-world GC pauses on large heaps.




Maximum HStoreFile size. If any one of a column families' HStoreFiles has grown to exceed this value, the hosting HRegion is split in two.




The time (in miliseconds) between 'major' compactions of all HStoreFiles in a region. Default: Set to 7 days. Major compactions tend to happen exactly when you need them least so enable them such that they run at off-peak for your deploy; or, since this setting is on a periodicity that is unlikely to match your loading, run the compactions via an external invocation out of a cron job or some such.




Jitter outer bound for major compactions. On each regionserver, we multiply the hbase.region.majorcompaction interval by some random fraction that is inside the bounds of this maximum. We then add this + or - product to when the next major compaction is to run. The idea is that major compaction does happen on every regionserver at exactly the same time. The smaller this number, the closer the compactions come together.




If more than this number of HStoreFiles in any one HStore (one HStoreFile is written per flush of memstore) then a compaction is run to rewrite all HStoreFiles files as one. Larger numbers put off compaction but when it runs, it takes longer to complete.




The number of flush threads. With less threads, the memstore flushes will be queued. With more threads, the flush will be executed in parallel, increasing the hdfs load. This can lead as well to more compactions.




If more than this number of StoreFiles in any one Store (one StoreFile is written per flush of MemStore) then updates are blocked for this HRegion until a compaction is completed, or until hbase.hstore.blockingWaitTime has been exceeded.




The time an HRegion will block updates for after hitting the StoreFile limit defined by hbase.hstore.blockingStoreFiles. After this time has elapsed, the HRegion will stop blocking updates even if a compaction has not been completed.




Max number of HStoreFiles to compact per 'minor' compaction.




How many KeyValues to read and then write in a batch when flushing or compacting. Do less if big KeyValues and problems with OOME. Do more if wide, small rows.




The amount of time to delay purging of delete markers with future timestamps. If unset, or set to 0, all delete markers, including those with future timestamps, are purged during the next major compaction. Otherwise, a delete marker is kept until the major compaction which occurs after the marker’s timestamp plus the value of this setting, in milliseconds.




Enables StoreFileScanner parallel-seeking in StoreScanner, a feature which can reduce response latency under special conditions.




The default thread pool size if parallel-seeking feature enabled.




Percentage of maximum heap (-Xmx setting) to allocate to block cache used by HFile/StoreFile. Default of 0.4 means allocate 40%. Set to 0 to disable but it’s not recommended; you need at least enough cache to hold the storefile indices.




This allows to put non-root multi-level index blocks into the block cache at the time the index is being written.




When the size of a leaf-level, intermediate-level, or root-level index block in a multi-level block index grows to this size, the block is written out and a new block is started.




Where to store the contents of the bucketcache. One of: heap, offheap, or file. If a file, set it to file:PATH_TO_FILE. See http://hbase.apache.org/book.html#offheap.blockcache for more information.




Whether or not the bucketcache is used in league with the LRU on-heap block cache. In this mode, indices and blooms are kept in the LRU blockcache and the data blocks are kept in the bucketcache.




A float that EITHER represents a percentage of total heap memory size to give to the cache (if < 1.0) OR, it is the total capacity in megabytes of BucketCache. Default: 0.0




A comma-separated list of sizes for buckets for the bucketcache. Can be multiple sizes. List block sizes in order from smallest to largest. The sizes you use will depend on your data access patterns. Must be a multiple of 1024 else you will run into 'java.io.IOException: Invalid HFile block magic' when you go to read from cache. If you specify no values here, then you pick up the default bucketsizes set in code (See BucketAllocator#DEFAULT_BUCKET_SIZES).




The HFile format version to use for new files. Version 3 adds support for tags in hfiles (See http://hbase.apache.org/book.html#hbase.tags). Distributed Log Replay requires that tags are enabled. Also see the configuration 'hbase.replication.rpc.codec'.




Enables cache-on-write for inline blocks of a compound Bloom filter.




The size in bytes of a single block ("chunk") of a compound Bloom filter. This size is approximate, because Bloom blocks can only be inserted at data block boundaries, and the number of keys per data block varies.




Whether an HFile block should be added to the block cache when the block is finished.




This is for the RPC layer to define how long (millisecond) HBase client applications take for a remote call to time out. It uses pings to check connections but will eventually throw a TimeoutException.




Operation timeout is a top-level restriction (millisecond) that makes sure a blocking operation in Table will not be blocked more than this. In each operation, if rpc request fails because of timeout or other reason, it will retry until success or throw RetriesExhaustedException. But if the total time being blocking reach the operation timeout before retries exhausted, it will break early and throw SocketTimeoutException.




The number of cells scanned in between heartbeat checks. Heartbeat checks occur during the processing of scans to determine whether or not the server should stop scanning in order to send back a heartbeat message to the client. Heartbeat messages are used to keep the client-server connection alive during long running scans. Small values mean that the heartbeat checks will occur more often and thus will provide a tighter bound on the execution time of the scan. Larger values mean that the heartbeat checks occur less frequently




This is another version of "hbase.rpc.timeout". For those RPC operation within cluster, we rely on this configuration to set a short timeout limitation for short operation. For example, short rpc timeout for region server’s trying to report to active master can benefit quicker master failover process.




Set no delay on rpc socket connections. See http://docs.oracle.com/javase/1.5.0/docs/api/java/net/Socket.html#getTcpNoDelay()




This config is for experts: don’t set its value unless you really know what you are doing. When set to a non-empty value, this represents the (external facing) hostname for the underlying server. See https://issues.apache.org/jira/browse/HBASE-12954 for details.




Full path to the kerberos keytab file to use for logging in the configured HMaster server principal.




Ex. "hbase/_HOST@EXAMPLE.COM". The kerberos principal name that should be used to run the HMaster process. The principal name should be in the form: user/hostname@DOMAIN. If "_HOST" is used as the hostname portion, it will be replaced with the actual hostname of the running instance.




Full path to the kerberos keytab file to use for logging in the configured HRegionServer server principal.




Ex. "hbase/_HOST@EXAMPLE.COM". The kerberos principal name that should be used to run the HRegionServer process. The principal name should be in the form: user/hostname@DOMAIN. If "_HOST" is used as the hostname portion, it will be replaced with the actual hostname of the running instance. An entry for this principal must exist in the file specified in hbase.regionserver.keytab.file




The policy configuration file used by RPC servers to make authorization decisions on client requests. Only used when HBase security is enabled.




List of users or groups (comma-separated), who are allowed full privileges, regardless of stored ACLs, across the cluster. Only used when HBase security is enabled.




The update interval for master key for authentication tokens in servers in milliseconds. Only used when HBase security is enabled.




The maximum lifetime in milliseconds after which an authentication token expires. Only used when HBase security is enabled.




When a client is configured to attempt a secure connection, but attempts to connect to an insecure server, that server may instruct the client to switch to SASL SIMPLE (unsecure) authentication. This setting controls whether or not the client will accept this instruction from the server. When false (the default), the client will not allow the fallback to SIMPLE authentication, and will abort the connection.




When a server is configured to require secure connections, it will reject connection attempts from clients using SASL SIMPLE (unsecure) authentication. This setting allows secure servers to accept SASL SIMPLE connections from clients when the client requests. When false (the default), the server will not allow the fallback to SIMPLE authentication, and will reject the connection. WARNING: This setting should ONLY be used as a temporary measure while converting clients over to secure authentication. It MUST BE DISABLED for secure operation.




Enables or disables coprocessor loading. If 'false' (disabled), any other coprocessor related configuration will be ignored.




Enables or disables user (aka. table) coprocessor loading. If 'false' (disabled), any table coprocessor attributes in table descriptors will be ignored. If "hbase.coprocessor.enabled" is 'false' this setting has no effect.




A comma-separated list of Coprocessors that are loaded by default on all tables. For any override coprocessor method, these classes will be called in order. After implementing your own Coprocessor, just put it in HBase’s classpath and add the fully qualified class name here. A coprocessor can also be loaded on demand by setting HTableDescriptor.




The port for the HBase REST server.




Defines the mode the REST server will be started in. Possible values are: false: All HTTP methods are permitted - GET/PUT/POST/DELETE. true: Only the GET method is permitted.




The maximum number of threads of the REST server thread pool. Threads in the pool are reused to process REST requests. This controls the maximum number of requests processed concurrently. It may help to control the memory used by the REST server to avoid OOM issues. If the thread pool is full, incoming requests will be queued up and wait for some free threads.




The minimum number of threads of the REST server thread pool. The thread pool always has at least these number of threads so the REST server is ready to serve incoming requests.




Enables running the REST server to support proxy-user mode.




Set to true to skip the 'hbase.defaults.for.version' check. Setting this to true can be useful in contexts other than the other side of a maven generation; i.e. running in an ide. You’ll want to set this boolean to true to avoid seeing the RuntimException complaint: "hbase-default.xml file seems to be for and old version of HBase (\${hbase.version}), this version is X.X.X-SNAPSHOT"




A comma-separated list of org.apache.hadoop.hbase.coprocessor.MasterObserver coprocessors that are loaded by default on the active HMaster process. For any implemented coprocessor methods, the listed classes will be called in order. After implementing your own MasterObserver, just put it in HBase’s classpath and add the fully qualified class name here.




Set to true to cause the hosting server (master or regionserver) to abort if a coprocessor fails to load, fails to initialize, or throws an unexpected Throwable object. Setting this to false will allow the server to continue execution but the system wide state of the coprocessor in question will become inconsistent as it will be properly executing in only a subset of servers, so this is most useful for debugging only.




Set true to enable online schema changes.




Set to true to enable locking the table in zookeeper for schema change operations. Table locking from master prevents concurrent schema modifications to corrupt table state.




Maximum size of single row in bytes (default is 1 Gb) for Get’ting or Scan’ning without in-row scan flag set. If row size exceeds this limit RowTooBigException is thrown to client.




The "core size" of the thread pool. New threads are created on every connection until this many threads are created.




The maximum size of the thread pool. When the pending request queue overflows, new threads are created until their number reaches this number. After that, the server starts dropping connections.




The maximum number of pending Thrift connections waiting in the queue. If there are no idle threads in the pool, the server queues requests. Only when the queue overflows, new threads are added, up to hbase.thrift.maxQueuedRequests threads.




The upper bound for the table pool used in the Thrift gateways server. Since this is per table name, we assume a single table and so with 1000 default worker threads max this is set to a matching number. For other workloads this number can be adjusted as needed.




Use Thrift TFramedTransport on the server side. This is the recommended transport for thrift servers and requires a similar setting on the client side. Changing this to false will select the default transport, vulnerable to DoS when malformed requests are issued due to THRIFT-601.




Default frame size when using framed transport




Use Thrift TCompactProtocol binary serialization protocol.




FS Permissions for the root directory in a secure(kerberos) setup. When master starts, it creates the rootdir with this permissions or sets the permissions if it does not match.




Enable, if true, that file permissions should be assigned to the files written by the regionserver




File permissions that should be used to write data files when hbase.data.umask.enable is true




Whether to include the prefix "tbl.tablename" in per-column family metrics. If true, for each metric M, per-cf metrics will be reported for tbl.T.cf.CF.M, if false, per-cf metrics will be aggregated by column-family across tables, and reported for cf.CF.M. In both cases, the aggregated metric M across tables and cfs will be reported.




Whether to report metrics about time taken performing an operation on the region server. Get, Put, Delete, Increment, and Append can all have their times exposed through Hadoop metrics per CF and per region.




Set to true to allow snapshots to be taken / restored / cloned.




Set to true to take a snapshot before the restore operation. The snapshot taken will be used in case of failure, to restore the previous state. At the end of the restore operation this snapshot will be deleted




Name of the failsafe snapshot taken by the restore operation. You can use the {snapshot.name}, {table.name} and {restore.timestamp} variables to create a name based on what you are restoring.




The number that determines how often we scan to see if compaction is necessary. Normally, compactions are done after some events (such as memstore flush), but if region didn’t receive a lot of writes for some time, or due to different compaction policies, it may be necessary to check it periodically. The interval between checks is hbase.server.compactchecker.interval.multiplier multiplied by hbase.server.thread.wakefrequency.




How long we wait on dfs lease recovery in total before giving up.




How long between dfs recover lease invocations. Should be larger than the sum of the time it takes for the namenode to issue a block recovery command as part of datanode; dfs.heartbeat.interval and the time it takes for the primary datanode, performing block recovery to timeout on a dead datanode; usually dfs.client.socket-timeout. See the end of HBASE-8389 for more.




New column family descriptors will use this value as the default number of versions to keep.




If the DFSClient configuration dfs.client.read.shortcircuit.buffer.size is unset, we will use what is configured here as the short circuit read default direct byte buffer size. DFSClient native default is 1MB; HBase keeps its HDFS files open so number of file blocks * 1MB soon starts to add up and threaten OOME because of a shortage of direct memory. So, we set it down from the default. Make it > the default hbase block size set in the HColumnDescriptor which is usually 64k.




If set to true (the default), HBase verifies the checksums for hfile blocks. HBase writes checksums inline with the data when it writes out hfiles. HDFS (as of this writing) writes checksums to a separate file than the data file necessitating extra seeks. Setting this flag saves some on i/o. Checksum verification by HDFS will be internally disabled on hfile streams when this flag is set. If the hbase-checksum verification fails, we will switch back to using HDFS checksums (so do not disable HDFS checksums! And besides this feature applies to hfiles only, not to WALs). If this parameter is set to false, then hbase will not verify any checksums, instead it will depend on checksum verification being done in the HDFS client.




Number of bytes in a newly created checksum chunk for HBase-level checksums in hfile blocks.




Name of an algorithm that is used to compute checksums. Possible values are NULL, CRC32, CRC32C.




Maximum number of bytes returned when calling a scanner’s next method. Note that when a single row is larger than this limit the row is still returned completely. The default value is 2MB, which is good for 1ge networks. With faster and/or high latency networks this value should be increased.




Maximum number of bytes returned when calling a scanner’s next method. Note that when a single row is larger than this limit the row is still returned completely. The default value is 100MB. This is a safety setting to protect the server from OOM situations.




This setting activates the publication by the master of the status of the region server. When a region server dies and its recovery starts, the master will push this information to the client application, to let them cut the connection immediately instead of waiting for a timeout.




Implementation of the status publication with a multicast message.




Implementation of the status listener with a multicast message.




Multicast address to use for the status publication by multicast.



Multicast port to use for the status publication by multicast.




The directory from which the custom filter/co-processor jars can be loaded dynamically by the region server without the need to restart. However, an already loaded filter/co-processor class would not be un-loaded. See HBASE-1936 for more details.




Controls whether or not secure authentication is enabled for HBase. Possible values are 'simple' (no authentication), and 'kerberos'.




Servlet filters for REST service.




Class used to execute the regions balancing when the period occurs. See the class comment for more on how it works http://hbase.apache.org/devapidocs/org/apache/hadoop/hbase/master/balancer/StochasticLoadBalancer.html It replaces the DefaultLoadBalancer as the default (since renamed as the SimpleLoadBalancer).




If this setting is enabled and ACL based access control is active (the AccessController coprocessor is installed either as a system coprocessor or on a table as a table coprocessor) then you must grant all relevant users EXEC privilege if they require the ability to execute coprocessor endpoint calls. EXEC privilege, like any other permission, can be granted globally to a user, or to a user on a per table or per namespace basis. For more information on coprocessor endpoints, see the coprocessor section of the HBase online manual. For more information on granting or revoking permissions using the AccessController, see the security section of the HBase online manual.




A comma-separated list of org.apache.hadoop.hbase.procedure.RegionServerProcedureManager procedure managers that are loaded by default on the active HRegionServer process. The lifecycle methods (init/start/stop) will be called by the active HRegionServer process to perform the specific globally barriered procedure. After implementing your own RegionServerProcedureManager, just put it in HBase’s classpath and add the fully qualified class name here.




A comma-separated list of org.apache.hadoop.hbase.procedure.MasterProcedureManager procedure managers that are loaded by default on the active HMaster process. A procedure is identified by its signature and users can use the signature and an instant name to trigger an execution of a globally barriered procedure. After implementing your own MasterProcedureManager, just put it in HBase’s classpath and add the fully qualified class name here.




Fully qualified name of class implementing coordinated state manager.




The period (in milliseconds) for refreshing the store files for the secondary regions. 0 means this feature is disabled. Secondary regions sees new files (from flushes and compactions) from primary once the secondary region refreshes the list of files in the region (there is no notification mechanism). But too frequent refreshes might cause extra Namenode pressure. If the files cannot be refreshed for longer than HFile TTL (hbase.master.hfilecleaner.ttl) the requests are rejected. Configuring HFile TTL to a larger value is also recommended with this setting.




Whether asynchronous WAL replication to the secondary region replicas is enabled or not. If this is enabled, a replication peer named "region_replica_replication" will be created which will tail the logs and replicate the mutatations to region replicas for tables that have region replication > 1. If this is enabled once, disabling this replication also requires disabling the replication peer using shell or ReplicationAdmin java class. Replication to secondary region replicas works over standard inter-cluster replication. So replication, if disabled explicitly, also has to be enabled by setting "hbase.replication" to true for this feature to work.




A comma separated list of class names. Each class in the list must extend org.apache.hadoop.hbase.http.FilterInitializer. The corresponding Filter will be initialized. Then, the Filter will be applied to all user facing jsp and servlet web pages. The ordering of the list defines the ordering of the filters. The default StaticUserWebFilter add a user principal as defined by the hbase.http.staticuser.user property.




This property if enabled, will check whether the labels in the visibility expression are associated with the user issuing the mutation




The maximum number of threads that the HTTP Server will create in its ThreadPool.




The codec that is to be used when replication is enabled so that the tags are also replicated. This is used along with HFileV3 which supports tags in them. If tags are not used or if the hfile version used is HFileV2 then KeyValueCodec can be used as the replication codec. Note that using KeyValueCodecWithTags for replication when there are no tags causes no harm.




The maximum number of threads any replication source will use for shipping edits to the sinks in parallel. This also limits the number of chunks each replication batch is broken into. Larger values can improve the replication throughput between the master and slave clusters. The default of 10 will rarely need to be changed.




The user name to filter as, on static web filters while rendering content. An example use is the HDFS web UI (user to be used for browsing files).




Class used to execute the region normalization when the period occurs. See the class comment for more on how it works http://hbase.apache.org/devapidocs/org/apache/hadoop/hbase/master/normalizer/SimpleRegionNormalizer.html




The percent of region server RPC threads failed to abort RS. -1 Disable aborting; 0 Abort if even a single handler has died; 0.x Abort only when this percent of handlers have died; 1 Abort only all of the handers have died.




Timeout for master for the snapshot procedure execution




Timeout for regionservers to keep threads in snapshot request pool waiting



7.3. hbase-env.sh

Set HBase environment variables in this file. Examples include options to pass the JVM on start of an HBase daemon such as heap size and garbage collector configs. You can also set configurations for HBase configuration, log directories, niceness, ssh options, where to locate process pid files, etc. Open the file at conf/hbase-env.sh and peruse its content. Each option is fairly well documented. Add your own environment variables here if you want them read by HBase daemons on startup.

Changes here will require a cluster restart for HBase to notice the change.

7.4. log4j.properties

Edit this file to change rate at which HBase files are rolled and to change the level at which HBase logs messages.

Changes here will require a cluster restart for HBase to notice the change though log levels can be changed for particular daemons via the HBase UI.

7.5. Client configuration and dependencies connecting to an HBase cluster

If you are running HBase in standalone mode, you don’t need to configure anything for your client to work provided that they are all on the same machine.

Since the HBase Master may move around, clients bootstrap by looking to ZooKeeper for current critical locations. ZooKeeper is where all these values are kept. Thus clients require the location of the ZooKeeper ensemble before they can do anything else. Usually this the ensemble location is kept out in the hbase-site.xml and is picked up by the client from the CLASSPATH.

If you are configuring an IDE to run an HBase client, you should include the conf/ directory on your classpath so hbase-site.xml settings can be found (or add src/test/resources to pick up the hbase-site.xml used by tests).

Minimally, a client of HBase needs several libraries in its CLASSPATH when connecting to a cluster, including:

commons-configuration (commons-configuration-1.6.jar)
commons-lang (commons-lang-2.5.jar)
commons-logging (commons-logging-1.1.1.jar)
hadoop-core (hadoop-core-1.0.0.jar)
hbase (hbase-0.92.0.jar)
log4j (log4j-1.2.16.jar)
slf4j-api (slf4j-api-1.5.8.jar)
slf4j-log4j (slf4j-log4j12-1.5.8.jar)
zookeeper (zookeeper-3.4.2.jar)

An example basic hbase-site.xml for client only might look as follows:

<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
    <description>The directory shared by region servers.

7.5.1. Java client configuration

The configuration used by a Java client is kept in an HBaseConfiguration instance.

The factory method on HBaseConfiguration, HBaseConfiguration.create();, on invocation, will read in the content of the first hbase-site.xml found on the client’s CLASSPATH, if one is present (Invocation will also factor in any hbase-default.xml found; an hbase-default.xml ships inside the hbase.X.X.X.jar). It is also possible to specify configuration directly without having to read from a hbase-site.xml. For example, to set the ZooKeeper ensemble for the cluster programmatically do as follows:

Configuration config = HBaseConfiguration.create();
config.set("hbase.zookeeper.quorum", "localhost");  // Here we are running zookeeper locally

If multiple ZooKeeper instances make up your ZooKeeper ensemble, they may be specified in a comma-separated list (just as in the hbase-site.xml file). This populated Configuration instance can then be passed to an Table, and so on.

7.6. Timeout settings

HBase provides a wide variety of timeout settings to limit the execution time of various remote operations.

  • hbase.rpc.timeout

  • hbase.client.operation.timeout

  • hbase.client.meta.operation.timeout

  • hbase.client.scanner.timeout.period

The hbase.rpc.timeout property limits how long a single RPC call can run before timing out.

A higher-level timeout is hbase.client.operation.timeout which is valid for each client call. When an RPC call fails for instance for a timeout due to hbase.rpc.timeout it will be retried until hbase.client.operation.timeout is reached. Client operation timeout for system tables can be fine tuned by setting hbase.client.meta.operation.timeout configuration value. When this is not set its value will use hbase.client.operation.timeout.

Timeout for scan operations is controlled differently. Use hbase.client.scanner.timeout.period property to set this timeout.

8. Example Configurations

8.1. Basic Distributed HBase Install

Here is an example basic configuration for a distributed ten node cluster: * The nodes are named example0, example1, etc., through node example9 in this example. * The HBase Master and the HDFS NameNode are running on the node example0. * RegionServers run on nodes example1-example9. * A 3-node ZooKeeper ensemble runs on example1, example2, and example3 on the default ports. * ZooKeeper data is persisted to the directory /export/zookeeper.

Below we show what the main configuration files — hbase-site.xml, regionservers, and hbase-env.sh — found in the HBase conf directory might look like.

8.1.1. hbase-site.xml

<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
    <description>The directory shared by RegionServers.
    <description>Property from ZooKeeper config zoo.cfg.
    The directory where the snapshot is stored.
    <description>The directory shared by RegionServers.
    <description>The mode the cluster will be in. Possible values are
      false: standalone and pseudo-distributed setups with managed ZooKeeper
      true: fully-distributed with unmanaged ZooKeeper Quorum (see hbase-env.sh)

8.1.2. regionservers

In this file you list the nodes that will run RegionServers. In our case, these nodes are example1-example9.


8.1.3. hbase-env.sh

The following lines in the hbase-env.sh file show how to set the JAVA_HOME environment variable (required for HBase 0.98.5 and newer) and set the heap to 4 GB (rather than the default value of 1 GB). If you copy and paste this example, be sure to adjust the JAVA_HOME to suit your environment.

# The java implementation to use.
export JAVA_HOME=/usr/java/jdk1.7.0/

# The maximum amount of heap to use. Default is left to JVM default.

Use rsync to copy the content of the conf directory to all nodes of the cluster.

9. The Important Configurations

Below we list some important configurations. We’ve divided this section into required configuration and worth-a-look recommended configs.

9.1. Required Configurations

Review the os and hadoop sections.

9.1.1. Big Cluster Configurations

If you have a cluster with a lot of regions, it is possible that a Regionserver checks in briefly after the Master starts while all the remaining RegionServers lag behind. This first server to check in will be assigned all regions which is not optimal. To prevent the above scenario from happening, up the hbase.master.wait.on.regionservers.mintostart property from its default value of 1. See HBASE-6389 Modify the conditions to ensure that Master waits for sufficient number of Region Servers before starting region assignments for more detail.

9.1.2. If a backup Master exists, make the primary Master fail fast

If the primary Master loses its connection with ZooKeeper, it will fall into a loop where it keeps trying to reconnect. Disable this functionality if you are running more than one Master: i.e. a backup Master. Failing to do so, the dying Master may continue to receive RPCs though another Master has assumed the role of primary. See the configuration fail.fast.expired.active.master.


The default timeout is three minutes (specified in milliseconds). This means that if a server crashes, it will be three minutes before the Master notices the crash and starts recovery. You might like to tune the timeout down to a minute or even less so the Master notices failures the sooner. Before changing this value, be sure you have your JVM garbage collection configuration under control otherwise, a long garbage collection that lasts beyond the ZooKeeper session timeout will take out your RegionServer (You might be fine with this — you probably want recovery to start on the server if a RegionServer has been in GC for a long period of time).

To change this configuration, edit hbase-site.xml, copy the changed file around the cluster and restart.

We set this value high to save our having to field questions up on the mailing lists asking why a RegionServer went down during a massive import. The usual cause is that their JVM is untuned and they are running into long GC pauses. Our thinking is that while users are getting familiar with HBase, we’d save them having to know all of its intricacies. Later when they’ve built some confidence, then they can play with configuration such as this.

Number of ZooKeeper Instances

See zookeeper.

9.2.2. HDFS Configurations


This is the "…​number of volumes that are allowed to fail before a DataNode stops offering service. By default any volume failure will cause a datanode to shutdown" from the hdfs-default.xml description. You might want to set this to about half the amount of your available disks.

9.2.3. hbase.regionserver.handler.count

This setting defines the number of threads that are kept open to answer incoming requests to user tables. The rule of thumb is to keep this number low when the payload per request approaches the MB (big puts, scans using a large cache) and high when the payload is small (gets, small puts, ICVs, deletes). The total size of the queries in progress is limited by the setting hbase.ipc.server.max.callqueue.size.

It is safe to set that number to the maximum number of incoming clients if their payload is small, the typical example being a cluster that serves a website since puts aren’t typically buffered and most of the operations are gets.

The reason why it is dangerous to keep this setting high is that the aggregate size of all the puts that are currently happening in a region server may impose too much pressure on its memory, or even trigger an OutOfMemoryError. A RegionServer running on low memory will trigger its JVM’s garbage collector to run more frequently up to a point where GC pauses become noticeable (the reason being that all the memory used to keep all the requests' payloads cannot be trashed, no matter how hard the garbage collector tries). After some time, the overall cluster throughput is affected since every request that hits that RegionServer will take longer, which exacerbates the problem even more.

You can get a sense of whether you have too little or too many handlers by rpc.logging on an individual RegionServer then tailing its logs (Queued requests consume memory).

9.2.4. Configuration for large memory machines

HBase ships with a reasonable, conservative configuration that will work on nearly all machine types that people might want to test with. If you have larger machines — HBase has 8G and larger heap — you might the following configuration options helpful. TODO.

9.2.5. Compression

You should consider enabling ColumnFamily compression. There are several options that are near-frictionless and in most all cases boost performance by reducing the size of StoreFiles and thus reducing I/O.

See compression for more information.

9.2.6. Configuring the size and number of WAL files

HBase uses wal to recover the memstore data that has not been flushed to disk in case of an RS failure. These WAL files should be configured to be slightly smaller than HDFS block (by default a HDFS block is 64Mb and a WAL file is ~60Mb).

HBase also has a limit on the number of WAL files, designed to ensure there’s never too much data that needs to be replayed during recovery. This limit needs to be set according to memstore configuration, so that all the necessary data would fit. It is recommended to allocate enough WAL files to store at least that much data (when all memstores are close to full). For example, with 16Gb RS heap, default memstore settings (0.4), and default WAL file size (~60Mb), 16Gb*0.4/60, the starting point for WAL file count is ~109. However, as all memstores are not expected to be full all the time, less WAL files can be allocated.

9.2.7. Managed Splitting

HBase generally handles splitting your regions, based upon the settings in your hbase-default.xml and hbase-site.xml configuration files. Important settings include hbase.regionserver.region.split.policy, hbase.hregion.max.filesize, hbase.regionserver.regionSplitLimit. A simplistic view of splitting is that when a region grows to hbase.hregion.max.filesize, it is split. For most use patterns, most of the time, you should use automatic splitting. See manual region splitting decisions for more information about manual region splitting.

Instead of allowing HBase to split your regions automatically, you can choose to manage the splitting yourself. This feature was added in HBase 0.90.0. Manually managing splits works if you know your keyspace well, otherwise let HBase figure where to split for you. Manual splitting can mitigate region creation and movement under load. It also makes it so region boundaries are known and invariant (if you disable region splitting). If you use manual splits, it is easier doing staggered, time-based major compactions to spread out your network IO load.

Disable Automatic Splitting

To disable automatic splitting, set hbase.hregion.max.filesize to a very large value, such as 100 GB It is not recommended to set it to its absolute maximum value of Long.MAX_VALUE.

Automatic Splitting Is Recommended

If you disable automatic splits to diagnose a problem or during a period of fast data growth, it is recommended to re-enable them when your situation becomes more stable. The potential benefits of managing region splits yourself are not undisputed.

Determine the Optimal Number of Pre-Split Regions

The optimal number of pre-split regions depends on your application and environment. A good rule of thumb is to start with 10 pre-split regions per server and watch as data grows over time. It is better to err on the side of too few regions and perform rolling splits later. The optimal number of regions depends upon the largest StoreFile in your region. The size of the largest StoreFile will increase with time if the amount of data grows. The goal is for the largest region to be just large enough that the compaction selection algorithm only compacts it during a timed major compaction. Otherwise, the cluster can be prone to compaction storms where a large number of regions under compaction at the same time. It is important to understand that the data growth causes compaction storms, and not the manual split decision.

If the regions are split into too many large regions, you can increase the major compaction interval by configuring HConstants.MAJOR_COMPACTION_PERIOD. HBase 0.90 introduced org.apache.hadoop.hbase.util.RegionSplitter, which provides a network-IO-safe rolling split of all regions.

9.2.8. Managed Compactions

By default, major compactions are scheduled to run once in a 7-day period. Prior to HBase 0.96.x, major compactions were scheduled to happen once per day by default.

If you need to control exactly when and how often major compaction runs, you can disable managed major compactions. See the entry for hbase.hregion.majorcompaction in the compaction.parameters table for details.

Do Not Disable Major Compactions

Major compactions are absolutely necessary for StoreFile clean-up. Do not disable them altogether. You can run major compactions manually via the HBase shell or via the Admin API.

For more information about compactions and the compaction file selection process, see compaction

9.2.9. Speculative Execution

Speculative Execution of MapReduce tasks is on by default, and for HBase clusters it is generally advised to turn off Speculative Execution at a system-level unless you need it for a specific case, where it can be configured per-job. Set the properties mapreduce.map.speculative and mapreduce.reduce.speculative to false.

9.3. Other Configurations

9.3.1. Balancer

The balancer is a periodic operation which is run on the master to redistribute regions on the cluster. It is configured via hbase.balancer.period and defaults to 300000 (5 minutes).

See master.processes.loadbalancer for more information on the LoadBalancer.

9.3.2. Disabling Blockcache

Do not turn off block cache (You’d do it by setting hbase.block.cache.size to zero). Currently we do not do well if you do this because the RegionServer will spend all its time loading HFile indices over and over again. If your working set is such that block cache does you no good, at least size the block cache such that HFile indices will stay up in the cache (you can get a rough idea on the size you need by surveying RegionServer UIs; you’ll see index block size accounted near the top of the webpage).

9.3.3. Nagle’s or the small package problem

If a big 40ms or so occasional delay is seen in operations against HBase, try the Nagles' setting. For example, see the user mailing list thread, Inconsistent scan performance with caching set to 1 and the issue cited therein where setting notcpdelay improved scan speeds. You might also see the graphs on the tail of HBASE-7008 Set scanner caching to a better default where our Lars Hofhansl tries various data sizes w/ Nagle’s on and off measuring the effect.

9.3.4. Better Mean Time to Recover (MTTR)

This section is about configurations that will make servers come back faster after a fail. See the Deveraj Das and Nicolas Liochon blog post Introduction to HBase Mean Time to Recover (MTTR) for a brief introduction.

The issue HBASE-8354 forces Namenode into loop with lease recovery requests is messy but has a bunch of good discussion toward the end on low timeouts and how to effect faster recovery including citation of fixes added to HDFS. Read the Varun Sharma comments. The below suggested configurations are Varun’s suggestions distilled and tested. Make sure you are running on a late-version HDFS so you have the fixes he refers too and himself adds to HDFS that help HBase MTTR (e.g. HDFS-3703, HDFS-3712, and HDFS-4791 — Hadoop 2 for sure has them and late Hadoop 1 has some). Set the following in the RegionServer.

  <description>How much time we allow elapse between calls to recover lease.
  Should be larger than the dfs timeout.</description>
  <description>Down the DFS timeout from 60 to 10 seconds.</description>

And on the NameNode/DataNode side, set the following to enable 'staleness' introduced in HDFS-3703, HDFS-3912.

  <description>Down the DFS timeout from 60 to 10 seconds.</description>
  <description>Down the DFS timeout from 8 * 60 to 10 seconds.</description>
  <description>Down from 60 seconds to 3.</description>
  <description>Down from 45 seconds to 3 (2 == 3 retries).</description>
  <description>Enable stale state in hdfs</description>
  <description>Down from default 30 seconds</description>
  <description>Enable stale state in hdfs</description>

9.3.5. JMX

JMX (Java Management Extensions) provides built-in instrumentation that enables you to monitor and manage the Java VM. To enable monitoring and management from remote systems, you need to set system property com.sun.management.jmxremote.port (the port number through which you want to enable JMX RMI connections) when you start the Java VM. See the official documentation for more information. Historically, besides above port mentioned, JMX opens two additional random TCP listening ports, which could lead to port conflict problem. (See HBASE-10289 for details)

As an alternative, You can use the coprocessor-based JMX implementation provided by HBase. To enable it in 0.99 or above, add below property in hbase-site.xml:

DO NOT set com.sun.management.jmxremote.port for Java VM at the same time.

Currently it supports Master and RegionServer Java VM. By default, the JMX listens on TCP port 10102, you can further configure the port using below properties:


The registry port can be shared with connector port in most cases, so you only need to configure regionserver.rmi.registry.port. However if you want to use SSL communication, the 2 ports must be configured to different values.

By default the password authentication and SSL communication is disabled. To enable password authentication, you need to update hbase-env.sh like below:

export HBASE_JMX_BASE="-Dcom.sun.management.jmxremote.authenticate=true                  \
                       -Dcom.sun.management.jmxremote.password.file=your_password_file   \


See example password/access file under $JRE_HOME/lib/management.

To enable SSL communication with password authentication, follow below steps:

#1. generate a key pair, stored in myKeyStore
keytool -genkey -alias jconsole -keystore myKeyStore

#2. export it to file jconsole.cert
keytool -export -alias jconsole -keystore myKeyStore -file jconsole.cert

#3. copy jconsole.cert to jconsole client machine, import it to jconsoleKeyStore
keytool -import -alias jconsole -keystore jconsoleKeyStore -file jconsole.cert

And then update hbase-env.sh like below:

export HBASE_JMX_BASE="-Dcom.sun.management.jmxremote.ssl=true                         \
                       -Djavax.net.ssl.keyStore=/home/tianq/myKeyStore                 \
                       -Djavax.net.ssl.keyStorePassword=your_password_in_step_1       \
                       -Dcom.sun.management.jmxremote.authenticate=true                \
                       -Dcom.sun.management.jmxremote.password.file=your_password file \


Finally start jconsole on the client using the key store:

jconsole -J-Djavax.net.ssl.trustStore=/home/tianq/jconsoleKeyStore
To enable the HBase JMX implementation on Master, you also need to add below property in hbase-site.xml:

The corresponding properties for port configuration are master.rmi.registry.port (by default 10101) and master.rmi.connector.port (by default the same as registry.port)

10. Dynamic Configuration

Since HBase 1.0.0, it is possible to change a subset of the configuration without requiring a server restart. In the HBase shell, there are new operators, update_config and update_all_config that will prompt a server or all servers to reload configuration.

Only a subset of all configurations can currently be changed in the running server. Here is an incomplete list: hbase.regionserver.thread.compaction.large, hbase.regionserver.thread.compaction.small, hbase.regionserver.thread.split, hbase.regionserver.thread.merge, as well as compaction policy and configurations and adjustment to offpeak hours. For the full list consult the patch attached to HBASE-12147 Porting Online Config Change from 89-fb.

11. Using Amazon S3 Storage

HBase is designed to be tightly coupled with HDFS, and testing of other filesystems has not been thorough.

The following limitations have been reported:

  • RegionServers should be deployed in Amazon EC2 to mitigate latency and bandwidth limitations when accessing the filesystem, and RegionServers must remain available to preserve data locality.

  • S3 writes each inbound and outbound file to disk, which adds overhead to each operation.

  • The best performance is achieved when all clients and servers are in the Amazon cloud, rather than a heterogenous architecture.

  • You must be aware of the location of hadoop.tmp.dir so that the local /tmp/ directory is not filled to capacity.

  • HBase has a different file usage pattern than MapReduce jobs and has been optimized for HDFS, rather than distant networked storage.

  • The s3a:// protocol is strongly recommended. The s3n:// and s3:// protocols have serious limitations and do not use the Amazon AWS SDK. The s3a:// protocol is supported for use with HBase if you use Hadoop 2.6.1 or higher with HBase 1.2 or higher. Hadoop 2.6.0 is not supported with HBase at all.

Configuration details for Amazon S3 and associated Amazon services such as EMR are out of the scope of the HBase documentation. See the Hadoop Wiki entry on Amazon S3 Storage and Amazon’s documentation for deploying HBase in EMR.

One use case that is well-suited for Amazon S3 is storing snapshots. See Storing Snapshots in an Amazon S3 Bucket.


You cannot skip major versions when upgrading. If you are upgrading from version 0.90.x to 0.94.x, you must first go from 0.90.x to 0.92.x and then go from 0.92.x to 0.94.x.

It may be possible to skip across versions — for example go from 0.92.2 straight to 0.98.0 just following the 0.96.x upgrade instructions — but these scenarios are untested.

Review Apache HBase Configuration, in particular Hadoop. Familiarize yourself with Support and Testing Expectations.

12. HBase version number and compatibility

HBase has two versioning schemes, pre-1.0 and post-1.0. Both are detailed below.

12.1. Post 1.0 versions

Starting with the 1.0.0 release, HBase is working towards Semantic Versioning for its release versioning. In summary:

Given a version number MAJOR.MINOR.PATCH, increment the:
  • MAJOR version when you make incompatible API changes,

  • MINOR version when you add functionality in a backwards-compatible manner, and

  • PATCH version when you make backwards-compatible bug fixes.

  • Additional labels for pre-release and build metadata are available as extensions to the MAJOR.MINOR.PATCH format.

Compatibility Dimensions

In addition to the usual API versioning considerations HBase has other compatibility dimensions that we need to consider.

Client-Server wire protocol compatibility
  • Allows updating client and server out of sync.

  • We could only allow upgrading the server first. I.e. the server would be backward compatible to an old client, that way new APIs are OK.

  • Example: A user should be able to use an old client to connect to an upgraded cluster.

Server-Server protocol compatibility
  • Servers of different versions can co-exist in the same cluster.

  • The wire protocol between servers is compatible.

  • Workers for distributed tasks, such as replication and log splitting, can co-exist in the same cluster.

  • Dependent protocols (such as using ZK for coordination) will also not be changed.

  • Example: A user can perform a rolling upgrade.

File format compatibility
  • Support file formats backward and forward compatible

  • Example: File, ZK encoding, directory layout is upgraded automatically as part of an HBase upgrade. User can rollback to the older version and everything will continue to work.

Client API compatibility
  • Allow changing or removing existing client APIs.

  • An API needs to deprecated for a major version before we will change/remove it.

  • APIs available in a patch version will be available in all later patch versions. However, new APIs may be added which will not be available in earlier patch versions.

  • Example: A user using a newly deprecated api does not need to modify application code with hbase api calls until the next major version.

Client Binary compatibility
  • Client code written to APIs available in a given patch release can run unchanged (no recompilation needed) against the new jars of later patch versions.

  • Client code written to APIs available in a given patch release might not run against the old jars from an earlier patch version.

  • Example: Old compiled client code will work unchanged with the new jars.

Server-Side Limited API compatibility (taken from Hadoop)
  • Internal APIs are marked as Stable, Evolving, or Unstable

  • This implies binary compatibility for coprocessors and plugins (pluggable classes, including replication) as long as these are only using marked interfaces/classes.

  • Example: Old compiled Coprocessor, Filter, or Plugin code will work unchanged with the new jars.

Dependency Compatibility
  • An upgrade of HBase will not require an incompatible upgrade of a dependent project, including the Java runtime.

  • Example: An upgrade of Hadoop will not invalidate any of the compatibilities guarantees we made.

Operational Compatibility
  • Metric changes

  • Behavioral changes of services

  • JMX APIs exposed via the /jmx/ endpoint

  • A patch upgrade is a drop-in replacement. Any change that is not Java binary compatible would not be allowed.[1]. Downgrading versions within patch releases may not be compatible.

  • A minor upgrade requires no application/client code modification. Ideally it would be a drop-in replacement but client code, coprocessors, filters, etc might have to be recompiled if new jars are used.

  • A major upgrade allows the HBase community to make breaking changes.

Table 3. Compatibility Matrix [2]




Client-Server wire Compatibility




Server-Server Compatibility




File Format Compatibility

N [3]



Client API Compatibility




Client Binary Compatibility




Server-Side Limited API Compatibility













Dependency Compatibility




Operational Compatibility




12.1.1. HBase API Surface

HBase has a lot of API points, but for the compatibility matrix above, we differentiate between Client API, Limited Private API, and Private API. HBase uses a version of Hadoop’s Interface classification. HBase’s Interface classification classes can be found here.

  • InterfaceAudience: captures the intended audience, possible values are Public (for end users and external projects), LimitedPrivate (for other Projects, Coprocessors or other plugin points), and Private (for internal use).

  • InterfaceStability: describes what types of interface changes are permitted. Possible values are Stable, Evolving, Unstable, and Deprecated.

HBase Client API

HBase Client API consists of all the classes or methods that are marked with InterfaceAudience.Public interface. All main classes in hbase-client and dependent modules have either InterfaceAudience.Public, InterfaceAudience.LimitedPrivate, or InterfaceAudience.Private marker. Not all classes in other modules (hbase-server, etc) have the marker. If a class is not annotated with one of these, it is assumed to be a InterfaceAudience.Private class.

HBase LimitedPrivate API

LimitedPrivate annotation comes with a set of target consumers for the interfaces. Those consumers are coprocessors, phoenix, replication endpoint implementations or similar. At this point, HBase only guarantees source and binary compatibility for these interfaces between patch versions.

HBase Private API

All classes annotated with InterfaceAudience.Private or all classes that do not have the annotation are for HBase internal use only. The interfaces and method signatures can change at any point in time. If you are relying on a particular interface that is marked Private, you should open a jira to propose changing the interface to be Public or LimitedPrivate, or an interface exposed for this purpose.

12.2. Pre 1.0 versions

Before the semantic versioning scheme pre-1.0, HBase tracked either Hadoop’s versions (0.2x) or 0.9x versions. If you are into the arcane, checkout our old wiki page on HBase Versioning which tries to connect the HBase version dots. Below sections cover ONLY the releases before 1.0.

Odd/Even Versioning or "Development" Series Releases

Ahead of big releases, we have been putting up preview versions to start the feedback cycle turning-over earlier. These "Development" Series releases, always odd-numbered, come with no guarantees, not even regards being able to upgrade between two sequential releases (we reserve the right to break compatibility across "Development" Series releases). Needless to say, these releases are not for production deploys. They are a preview of what is coming in the hope that interested parties will take the release for a test drive and flag us early if we there are issues we’ve missed ahead of our rolling a production-worthy release.

Our first "Development" Series was the 0.89 set that came out ahead of HBase 0.90.0. HBase 0.95 is another "Development" Series that portends HBase 0.96.0. 0.99.x is the last series in "developer preview" mode before 1.0. Afterwards, we will be using semantic versioning naming scheme (see above).

Binary Compatibility

When we say two HBase versions are compatible, we mean that the versions are wire and binary compatible. Compatible HBase versions means that clients can talk to compatible but differently versioned servers. It means too that you can just swap out the jars of one version and replace them with the jars of another, compatible version and all will just work. Unless otherwise specified, HBase point versions are (mostly) binary compatible. You can safely do rolling upgrades between binary compatible versions; i.e. across point versions: e.g. from 0.94.5 to 0.94.6. See link:[Does compatibility between versions also mean binary compatibility?] discussion on the HBase dev mailing list.

12.3. Rolling Upgrades

A rolling upgrade is the process by which you update the servers in your cluster a server at a time. You can rolling upgrade across HBase versions if they are binary or wire compatible. See Rolling Upgrade Between Versions that are Binary/Wire Compatible for more on what this means. Coarsely, a rolling upgrade is a graceful stop each server, update the software, and then restart. You do this for each server in the cluster. Usually you upgrade the Master first and then the RegionServers. See Rolling Restart for tools that can help use the rolling upgrade process.

For example, in the below, HBase was symlinked to the actual HBase install. On upgrade, before running a rolling restart over the cluster, we changed the symlink to point at the new HBase software version and then ran

$ HADOOP_HOME=~/hadoop-2.6.0-CRC-SNAPSHOT ~/hbase/bin/rolling-restart.sh --config ~/conf_hbase

The rolling-restart script will first gracefully stop and restart the master, and then each of the RegionServers in turn. Because the symlink was changed, on restart the server will come up using the new HBase version. Check logs for errors as the rolling upgrade proceeds.

Rolling Upgrade Between Versions that are Binary/Wire Compatible

Unless otherwise specified, HBase point versions are binary compatible. You can do a Rolling Upgrades between HBase point versions. For example, you can go to 0.94.6 from 0.94.5 by doing a rolling upgrade across the cluster replacing the 0.94.5 binary with a 0.94.6 binary.

In the minor version-particular sections below, we call out where the versions are wire/protocol compatible and in this case, it is also possible to do a Rolling Upgrades. For example, in Rolling upgrade from 0.98.x to HBase 1.0.0, we state that it is possible to do a rolling upgrade between hbase-0.98.x and hbase-1.0.0.

13. Upgrade Paths

13.1. Upgrading from 0.98.x to 1.x

In this section we first note the significant changes that come in with 1.0.0+ HBase and then we go over the upgrade process. Be sure to read the significant changes section with care so you avoid surprises.

13.1.1. Changes of Note!

In here we list important changes that are in 1.0.0+ since 0.98.x., changes you should be aware that will go into effect once you upgrade.

ZooKeeper 3.4 is required in HBase 1.0.0+

See ZooKeeper Requirements.

HBase Default Ports Changed

The ports used by HBase changed. They used to be in the 600XX range. In HBase 1.0.0 they have been moved up out of the ephemeral port range and are 160XX instead (Master web UI was 60010 and is now 16010; the RegionServer web UI was 60030 and is now 16030, etc.). If you want to keep the old port locations, copy the port setting configs from hbase-default.xml into hbase-site.xml, change them back to the old values from the HBase 0.98.x era, and ensure you’ve distributed your configurations before you restart.

HBase Master Port Binding Change

In HBase 1.0.x, the HBase Master binds the RegionServer ports as well as the Master ports. This behavior is changed from HBase versions prior to 1.0. In HBase 1.1 and 2.0 branches, this behavior is reverted to the pre-1.0 behavior of the HBase master not binding the RegionServer ports.

hbase.bucketcache.percentage.in.combinedcache configuration has been REMOVED

You may have made use of this configuration if you are using BucketCache. If NOT using BucketCache, this change does not affect you. Its removal means that your L1 LruBlockCache is now sized using hfile.block.cache.size — i.e. the way you would size the on-heap L1 LruBlockCache if you were NOT doing BucketCache — and the BucketCache size is not whatever the setting for hbase.bucketcache.size is. You may need to adjust configs to get the LruBlockCache and BucketCache sizes set to what they were in 0.98.x and previous. If you did not set this config., its default value was 0.9. If you do nothing, your BucketCache will increase in size by 10%. Your L1 LruBlockCache will become hfile.block.cache.size times your java heap size (hfile.block.cache.size is a float between 0.0 and 1.0). To read more, see HBASE-11520 Simplify offheap cache config by removing the confusing "hbase.bucketcache.percentage.in.combinedcache".

If you have your own customer filters.

See the release notes on the issue HBASE-12068 [Branch-1] Avoid need to always do KeyValueUtil#ensureKeyValue for Filter transformCell; be sure to follow the recommendations therein.

Distributed Log Replay

Distributed Log Replay is off by default in HBase 1.0.0. Enabling it can make a big difference improving HBase MTTR. Enable this feature if you are doing a clean stop/start when you are upgrading. You cannot rolling upgrade to this feature (caveat if you are running on a version of HBase in excess of HBase 0.98.4 — see HBASE-12577 Disable distributed log replay by default for more).

Mismatch Of hbase.client.scanner.max.result.size Between Client and Server

If either the client or server version is lower than 0.98.11/1.0.0 and the server has a smaller value for hbase.client.scanner.max.result.size than the client, scan requests that reach the server’s hbase.client.scanner.max.result.size are likely to miss data. In particular, 0.98.11 defaults hbase.client.scanner.max.result.size to 2 MB but other versions default to larger values. For this reason, be very careful using 0.98.11 servers with any other client version.

Availability of Date Tiered Compaction.

The Date Tiered Compaction feature available as of 0.98.19 is available in the 1.y release line starting in release 1.3.0. If you have enabled this feature for any tables you must upgrade to version 1.3.0 or later. If you attempt to use an earlier 1.y release, any tables configured to use date tiered compaction will fail to have their regions open.

13.1.2. Rolling upgrade from 0.98.x to HBase 1.0.0

From 0.96.x to 1.0.0
You cannot do a rolling upgrade from 0.96.x to 1.0.0 without first doing a rolling upgrade to 0.98.x. See comment in HBASE-11164 Document and test rolling updates from 0.98 → 1.0 for the why. Also because HBase 1.0.0 enables HFile v3 by default, HBASE-9801 Change the default HFile version to V3, and support for HFile v3 only arrives in 0.98, this is another reason you cannot rolling upgrade from HBase 0.96.x; if the rolling upgrade stalls, the 0.96.x servers cannot open files written by the servers running the newer HBase 1.0.0 with HFile’s of version 3.

There are no known issues running a rolling upgrade from HBase 0.98.x to HBase 1.0.0.

13.1.3. Upgrading to 1.0 from 0.94

You cannot rolling upgrade from 0.94.x to 1.x.x. You must stop your cluster, install the 1.x.x software, run the migration described at Executing the 0.96 Upgrade (substituting 1.x.x. wherever we make mention of 0.96.x in the section below), and then restart. Be sure to upgrade your ZooKeeper if it is a version less than the required 3.4.x.

13.2. Upgrading from 0.96.x to 0.98.x

A rolling upgrade from 0.96.x to 0.98.x works. The two versions are not binary compatible.

Additional steps are required to take advantage of some of the new features of 0.98.x, including cell visibility labels, cell ACLs, and transparent server side encryption. See Securing Apache HBase for more information. Significant performance improvements include a change to the write ahead log threading model that provides higher transaction throughput under high load, reverse scanners, MapReduce over snapshot files, and striped compaction.

Clients and servers can run with 0.98.x and 0.96.x versions. However, applications may need to be recompiled due to changes in the Java API.

13.3. Upgrading from 0.94.x to 0.98.x

A rolling upgrade from 0.94.x directly to 0.98.x does not work. The upgrade path follows the same procedures as Upgrading from 0.94.x to 0.96.x. Additional steps are required to use some of the new features of 0.98.x. See Upgrading from 0.96.x to 0.98.x for an abbreviated list of these features.

13.4. Upgrading from 0.94.x to 0.96.x

13.4.1. The "Singularity"

HBase 0.96.x was EOL’d, September 1st, 2014
Do not deploy 0.96.x Deploy at least 0.98.x. See EOL 0.96.

You will have to stop your old 0.94.x cluster completely to upgrade. If you are replicating between clusters, both clusters will have to go down to upgrade. Make sure it is a clean shutdown. The less WAL files around, the faster the upgrade will run (the upgrade will split any log files it finds in the filesystem as part of the upgrade process). All clients must be upgraded to 0.96 too.

The API has changed. You will need to recompile your code against 0.96 and you may need to adjust applications to go against new APIs (TODO: List of changes).

13.4.2. Executing the 0.96 Upgrade

HDFS and ZooKeeper must be up!
HDFS and ZooKeeper should be up and running during the upgrade process.

HBase 0.96.0 comes with an upgrade script. Run

$ bin/hbase upgrade

to see its usage. The script has two main modes: -check, and -execute.


The check step is run against a running 0.94 cluster. Run it from a downloaded 0.96.x binary. The check step is looking for the presence of HFile v1 files. These are unsupported in HBase 0.96.0. To have them rewritten as HFile v2 you must run a compaction.

The check step prints stats at the end of its run (grep for “Result:” in the log) printing absolute path of the tables it scanned, any HFile v1 files found, the regions containing said files (these regions will need a major compaction), and any corrupted files if found. A corrupt file is unreadable, and so is undefined (neither HFile v1 nor HFile v2).

To run the check step, run

$ bin/hbase upgrade -check

Here is sample output:

Tables Processed:

Count of HFileV1: 2
hdfs://localhost:41020/myHBase/usertable    /fa02dac1f38d03577bd0f7e666f12812/family/249450144068442524
hdfs://localhost:41020/myHBase/usertable    /ecdd3eaee2d2fcf8184ac025555bb2af/family/249450144068442512

Count of corrupted files: 1
Corrupted Files:
Count of Regions with HFileV1: 2
Regions to Major Compact:

There are some HFileV1, or corrupt files (files with incorrect major version)

In the above sample output, there are two HFile v1 files in two regions, and one corrupt file. Corrupt files should probably be removed. The regions that have HFile v1s need to be major compacted. To major compact, start up the hbase shell and review how to compact an individual region. After the major compaction is done, rerun the check step and the HFile v1 files should be gone, replaced by HFile v2 instances.

By default, the check step scans the HBase root directory (defined as hbase.rootdir in the configuration). To scan a specific directory only, pass the -dir option.

$ bin/hbase upgrade -check -dir /myHBase/testTable

The above command would detect HFile v1 files in the /myHBase/testTable directory.

Once the check step reports all the HFile v1 files have been rewritten, it is safe to proceed with the upgrade.


After the check step shows the cluster is free of HFile v1, it is safe to proceed with the upgrade. Next is the execute step. You must SHUTDOWN YOUR 0.94.x CLUSTER before you can run the execute step. The execute step will not run if it detects running HBase masters or RegionServers.

HDFS and ZooKeeper should be up and running during the upgrade process. If zookeeper is managed by HBase, then you can start zookeeper so it is available to the upgrade by running

$ ./hbase/bin/hbase-daemon.sh start zookeeper

The execute upgrade step is made of three substeps.

  • Namespaces: HBase 0.96.0 has support for namespaces. The upgrade needs to reorder directories in the filesystem for namespaces to work.

  • ZNodes: All znodes are purged so that new ones can be written in their place using a new protobuf’ed format and a few are migrated in place: e.g. replication and table state znodes

  • WAL Log Splitting: If the 0.94.x cluster shutdown was not clean, we’ll split WAL logs as part of migration before we startup on 0.96.0. This WAL splitting runs slower than the native distributed WAL splitting because it is all inside the single upgrade process (so try and get a clean shutdown of the 0.94.0 cluster if you can).

To run the execute step, make sure that first you have copied HBase 0.96.0 binaries everywhere under servers and under clients. Make sure the 0.94.0 cluster is down. Then do as follows:

$ bin/hbase upgrade -execute

Here is some sample output.

Starting Namespace upgrade
Created version file at hdfs://localhost:41020/myHBase with version=7
Migrating table testTable to hdfs://localhost:41020/myHBase/.data/default/testTable
Created version file at hdfs://localhost:41020/myHBase with version=8
Successfully completed NameSpace upgrade.
Starting Znode upgrade
Successfully completed Znode upgrade

Starting Log splitting
Successfully completed Log splitting

If the output from the execute step looks good, stop the zookeeper instance you started to do the upgrade:

$ ./hbase/bin/hbase-daemon.sh stop zookeeper

Now start up hbase-0.96.0.

13.5. Troubleshooting

Old Client connecting to 0.96 cluster

It will fail with an exception like the below. Upgrade.

17:22:15  Exception in thread "main" java.lang.IllegalArgumentException: Not a host:port pair: PBUF
17:22:15  *
17:22:15   api-compat-8.ent.cloudera.com ��  ���(
17:22:15    at org.apache.hadoop.hbase.util.Addressing.parseHostname(Addressing.java:60)
17:22:15    at org.apache.hadoop.hbase.ServerName.&init>(ServerName.java:101)
17:22:15    at org.apache.hadoop.hbase.ServerName.parseVersionedServerName(ServerName.java:283)
17:22:15    at org.apache.hadoop.hbase.MasterAddressTracker.bytesToServerName(MasterAddressTracker.java:77)
17:22:15    at org.apache.hadoop.hbase.MasterAddressTracker.getMasterAddress(MasterAddressTracker.java:61)
17:22:15    at org.apache.hadoop.hbase.client.HConnectionManager$HConnectionImplementation.getMaster(HConnectionManager.java:703)
17:22:15    at org.apache.hadoop.hbase.client.HBaseAdmin.&init>(HBaseAdmin.java:126)
17:22:15    at Client_4_3_0.setup(Client_4_3_0.java:716)
17:22:15    at Client_4_3_0.main(Client_4_3_0.java:63)

13.5.1. Upgrading META to use Protocol Buffers (Protobuf)

When you upgrade from versions prior to 0.96, META needs to be converted to use protocol buffers. This is controlled by the configuration option hbase.MetaMigrationConvertingToPB, which is set to true by default. Therefore, by default, no action is required on your part.

The migration is a one-time event. However, every time your cluster starts, META is scanned to ensure that it does not need to be converted. If you have a very large number of regions, this scan can take a long time. Starting in 0.98.5, you can set hbase.MetaMigrationConvertingToPB to false in hbase-site.xml, to disable this start-up scan. This should be considered an expert-level setting.

13.6. Upgrading from 0.92.x to 0.94.x

We used to think that 0.92 and 0.94 were interface compatible and that you can do a rolling upgrade between these versions but then we figured that HBASE-5357 Use builder pattern in HColumnDescriptor changed method signatures so rather than return void they instead return HColumnDescriptor. This will throw java.lang.NoSuchMethodError: org.apache.hadoop.hbase.HColumnDescriptor.setMaxVersions(I)V so 0.92 and 0.94 are NOT compatible. You cannot do a rolling upgrade between them.

13.7. Upgrading from 0.90.x to 0.92.x

13.7.1. Upgrade Guide

You will find that 0.92.0 runs a little differently to 0.90.x releases. Here are a few things to watch out for upgrading from 0.90.x to 0.92.0.


These are the important things to know before upgrading. . Once you upgrade, you can’t go back.

  1. MSLAB is on by default. Watch that heap usage if you have a lot of regions.

  2. Distributed Log Splitting is on by default. It should make RegionServer failover faster.

  3. There’s a separate tarball for security.

  4. If -XX:MaxDirectMemorySize is set in your hbase-env.sh, it’s going to enable the experimental off-heap cache (You may not want this).

You can’t go back!

To move to 0.92.0, all you need to do is shutdown your cluster, replace your HBase 0.90.x with HBase 0.92.0 binaries (be sure you clear out all 0.90.x instances) and restart (You cannot do a rolling restart from 0.90.x to 0.92.x — you must restart). On startup, the .META. table content is rewritten removing the table schema from the info:regioninfo column. Also, any flushes done post first startup will write out data in the new 0.92.0 file format, HBase file format with inline blocks (version 2). This means you cannot go back to 0.90.x once you’ve started HBase 0.92.0 over your HBase data directory.

MSLAB is ON by default

In 0.92.0, the hbase.hregion.memstore.mslab.enabled flag is set to true (See Long GC pauses). In 0.90.x it was false. When it is enabled, memstores will step allocate memory in MSLAB 2MB chunks even if the memstore has zero or just a few small elements. This is fine usually but if you had lots of regions per RegionServer in a 0.90.x cluster (and MSLAB was off), you may find yourself OOME’ing on upgrade because the thousands of regions * number of column families * 2MB MSLAB (at a minimum) puts your heap over the top. Set hbase.hregion.memstore.mslab.enabled to false or set the MSLAB size down from 2MB by setting hbase.hregion.memstore.mslab.chunksize to something less.

Distributed Log Splitting is on by default

Previous, WAL logs on crash were split by the Master alone. In 0.92.0, log splitting is done by the cluster (See HBASE-1364 [performance] Distributed splitting of regionserver commit logs or see the blog post Apache HBase Log Splitting). This should cut down significantly on the amount of time it takes splitting logs and getting regions back online again.

Memory accounting is different now

In 0.92.0, HBase file format with inline blocks (version 2) indices and bloom filters take up residence in the same LRU used caching blocks that come from the filesystem. In 0.90.x, the HFile v1 indices lived outside of the LRU so they took up space even if the index was on a ‘cold’ file, one that wasn’t being actively used. With the indices now in the LRU, you may find you have less space for block caching. Adjust your block cache accordingly. See the Block Cache for more detail. The block size default size has been changed in 0.92.0 from 0.2 (20 percent of heap) to 0.25.

On the Hadoop version to use

Run 0.92.0 on Hadoop 1.0.x (or CDH3u3). The performance benefits are worth making the move. Otherwise, our Hadoop prescription is as it has been; you need an Hadoop that supports a working sync. See Hadoop.

If running on Hadoop 1.0.x (or CDH3u3), enable local read. See Practical Caching presentation for ruminations on the performance benefits ‘going local’ (and for how to enable local reads).

HBase 0.92.0 ships with ZooKeeper 3.4.2

If you can, upgrade your ZooKeeper. If you can’t, 3.4.2 clients should work against 3.3.X ensembles (HBase makes use of 3.4.2 API).

Online alter is off by default

In 0.92.0, we’ve added an experimental online schema alter facility (See hbase.online.schema.update.enable). It’s off by default. Enable it at your own risk. Online alter and splitting tables do not play well together so be sure your cluster quiescent using this feature (for now).


The web UI has had a few additions made in 0.92.0. It now shows a list of the regions currently transitioning, recent compactions/flushes, and a process list of running processes (usually empty if all is well and requests are being handled promptly). Other additions including requests by region, a debugging servlet dump, etc.

Security tarball

We now ship with two tarballs; secure and insecure HBase. Documentation on how to setup a secure HBase is on the way.

Changes in HBase replication

0.92.0 adds two new features: multi-slave and multi-master replication. The way to enable this is the same as adding a new peer, so in order to have multi-master you would just run add_peer for each cluster that acts as a master to the other slave clusters. Collisions are handled at the timestamp level which may or may not be what you want, this needs to be evaluated on a per use case basis. Replication is still experimental in 0.92 and is disabled by default, run it at your own risk.

RegionServer now aborts if OOME

If an OOME, we now have the JVM kill -9 the RegionServer process so it goes down fast. Previous, a RegionServer might stick around after incurring an OOME limping along in some wounded state. To disable this facility, and recommend you leave it in place, you’d need to edit the bin/hbase file. Look for the addition of the -XX:OnOutOfMemoryError="kill -9 %p" arguments (See HBASE-4769 - ‘Abort RegionServer Immediately on OOME’).

HFile v2 and the “Bigger, Fewer” Tendency

0.92.0 stores data in a new format, HBase file format with inline blocks (version 2). As HBase runs, it will move all your data from HFile v1 to HFile v2 format. This auto-migration will run in the background as flushes and compactions run. HFile v2 allows HBase run with larger regions/files. In fact, we encourage that all HBasers going forward tend toward Facebook axiom #1, run with larger, fewer regions. If you have lots of regions now — more than 100s per host — you should look into setting your region size up after you move to 0.92.0 (In 0.92.0, default size is now 1G, up from 256M), and then running online merge tool (See HBASE-1621 merge tool should work on online cluster, but disabled table).

13.8. Upgrading to HBase 0.90.x from 0.20.x or 0.89.x

This version of 0.90.x HBase can be started on data written by HBase 0.20.x or HBase 0.89.x. There is no need of a migration step. HBase 0.89.x and 0.90.x does write out the name of region directories differently — it names them with a md5 hash of the region name rather than a jenkins hash — so this means that once started, there is no going back to HBase 0.20.x.

Be sure to remove the hbase-default.xml from your conf directory on upgrade. A 0.20.x version of this file will have sub-optimal configurations for 0.90.x HBase. The hbase-default.xml file is now bundled into the HBase jar and read from there. If you would like to review the content of this file, see it in the src tree at src/main/resources/hbase-default.xml or see HBase Default Configuration.

Finally, if upgrading from 0.20.x, check your .META. schema in the shell. In the past we would recommend that users run with a 16kb MEMSTORE_FLUSHSIZE. Run

hbase> scan '-ROOT-'

in the shell. This will output the current .META. schema. Check MEMSTORE_FLUSHSIZE size. Is it 16kb (16384)? If so, you will need to change this (The 'normal'/default value is 64MB (67108864)). Run the script bin/set_meta_memstore_size.rb. This will make the necessary edit to your .META. schema. Failure to run this change will make for a slow cluster. See HBASE-3499 Users upgrading to 0.90.0 need to have their .META. table updated with the right MEMSTORE_SIZE.

The Apache HBase Shell

The Apache HBase Shell is (J)Ruby's IRB with some HBase particular commands added. Anything you can do in IRB, you should be able to do in the HBase Shell.

To run the HBase shell, do as follows:

$ ./bin/hbase shell

Type help and then <RETURN> to see a listing of shell commands and options. Browse at least the paragraphs at the end of the help output for the gist of how variables and command arguments are entered into the HBase shell; in particular note how table names, rows, and columns, etc., must be quoted.

See shell exercises for example basic shell operation.

Here is a nicely formatted listing of all shell commands by Rajeshbabu Chintaguntla.

14. Scripting with Ruby

For examples scripting Apache HBase, look in the HBase bin directory. Look at the files that end in *.rb. To run one of these files, do as follows:

$ ./bin/hbase org.jruby.Main PATH_TO_SCRIPT

15. Running the Shell in Non-Interactive Mode

A new non-interactive mode has been added to the HBase Shell (HBASE-11658). Non-interactive mode captures the exit status (success or failure) of HBase Shell commands and passes that status back to the command interpreter. If you use the normal interactive mode, the HBase Shell will only ever return its own exit status, which will nearly always be 0 for success.

To invoke non-interactive mode, pass the -n or --non-interactive option to HBase Shell.

16. HBase Shell in OS Scripts

You can use the HBase shell from within operating system script interpreters like the Bash shell which is the default command interpreter for most Linux and UNIX distributions. The following guidelines use Bash syntax, but could be adjusted to work with C-style shells such as csh or tcsh, and could probably be modified to work with the Microsoft Windows script interpreter as well. Submissions are welcome.

Spawning HBase Shell commands in this way is slow, so keep that in mind when you are deciding when combining HBase operations with the operating system command line is appropriate.
Example 8. Passing Commands to the HBase Shell

You can pass commands to the HBase Shell in non-interactive mode (see hbase.shell.noninteractive) using the echo command and the | (pipe) operator. Be sure to escape characters in the HBase commands which would otherwise be interpreted by the shell. Some debug-level output has been truncated from the example below.

$ echo "describe 'test1'" | ./hbase shell -n

Version 0.98.3-hadoop2, rd5e65a9144e315bb0a964e7730871af32f5018d5, Sat May 31 19:56:09 PDT 2014

describe 'test1'

DESCRIPTION                                          ENABLED
 'test1', {NAME => 'cf', DATA_BLOCK_ENCODING => 'NON true
 'false', BLOCKSIZE => '65536', IN_MEMORY => 'false'
 , BLOCKCACHE => 'true'}
1 row(s) in 3.2410 seconds

To suppress all output, echo it to /dev/null:

$ echo "describe 'test'" | ./hbase shell -n > /dev/null 2>&1
Example 9. Checking the Result of a Scripted Command

Since scripts are not designed to be run interactively, you need a way to check whether your command failed or succeeded. The HBase shell uses the standard convention of returning a value of 0 for successful commands, and some non-zero value for failed commands. Bash stores a command’s return value in a special environment variable called $?. Because that variable is overwritten each time the shell runs any command, you should store the result in a different, script-defined variable.

This is a naive script that shows one way to store the return value and make a decision based upon it.


echo "describe 'test'" | ./hbase shell -n > /dev/null 2>&1
echo "The status was " $status
if ($status == 0); then
    echo "The command succeeded"
    echo "The command may have failed."
return $status

16.1. Checking for Success or Failure In Scripts

Getting an exit code of 0 means that the command you scripted definitely succeeded. However, getting a non-zero exit code does not necessarily mean the command failed. The command could have succeeded, but the client lost connectivity, or some other event obscured its success. This is because RPC commands are stateless. The only way to be sure of the status of an operation is to check. For instance, if your script creates a table, but returns a non-zero exit value, you should check whether the table was actually created before trying again to create it.

17. Read HBase Shell Commands from a Command File

You can enter HBase Shell commands into a text file, one command per line, and pass that file to the HBase Shell.

Example 10. Example Command File
create 'test', 'cf'
list 'test'
put 'test', 'row1', 'cf:a', 'value1'
put 'test', 'row2', 'cf:b', 'value2'
put 'test', 'row3', 'cf:c', 'value3'
put 'test', 'row4', 'cf:d', 'value4'
scan 'test'
get 'test', 'row1'
disable 'test'
enable 'test'
Example 11. Directing HBase Shell to Execute the Commands

Pass the path to the command file as the only argument to the hbase shell command. Each command is executed and its output is shown. If you do not include the exit command in your script, you are returned to the HBase shell prompt. There is no way to programmatically check each individual command for success or failure. Also, though you see the output for each command, the commands themselves are not echoed to the screen so it can be difficult to line up the command with its output.

$ ./hbase shell ./sample_commands.txt
0 row(s) in 3.4170 seconds

1 row(s) in 0.0590 seconds

0 row(s) in 0.1540 seconds

0 row(s) in 0.0080 seconds

0 row(s) in 0.0060 seconds

0 row(s) in 0.0060 seconds

ROW                   COLUMN+CELL
 row1                 column=cf:a, timestamp=1407130286968, value=value1
 row2                 column=cf:b, timestamp=1407130286997, value=value2
 row3                 column=cf:c, timestamp=1407130287007, value=value3
 row4                 column=cf:d, timestamp=1407130287015, value=value4
4 row(s) in 0.0420 seconds

COLUMN                CELL
 cf:a                 timestamp=1407130286968, value=value1
1 row(s) in 0.0110 seconds

0 row(s) in 1.5630 seconds

0 row(s) in 0.4360 seconds

18. Passing VM Options to the Shell

You can pass VM options to the HBase Shell using the HBASE_SHELL_OPTS environment variable. You can set this in your environment, for instance by editing ~/.bashrc, or set it as part of the command to launch HBase Shell. The following example sets several garbage-collection-related variables, just for the lifetime of the VM running the HBase Shell. The command should be run all on a single line, but is broken by the \ character, for readability.

$ HBASE_SHELL_OPTS="-verbose:gc -XX:+PrintGCApplicationStoppedTime -XX:+PrintGCDateStamps \
  -XX:+PrintGCDetails -Xloggc:$HBASE_HOME/logs/gc-hbase.log" ./bin/hbase shell

19. Overriding configuration starting the HBase Shell

As of hbase-2.0.5/hbase-2.1.3/hbase-2.2.0/hbase-1.4.10/hbase-1.5.0, you can pass or override hbase configuration as specified in hbase-*.xml by passing your key/values prefixed with -D on the command-line as follows:

$ ./bin/hbase shell -Dhbase.zookeeper.quorum=ZK0.remote.cluster.example.org,ZK1.remote.cluster.example.org,ZK2.remote.cluster.example.org -Draining=false
hbase(main):001:0> @shell.hbase.configuration.get("hbase.zookeeper.quorum")
=> "ZK0.remote.cluster.example.org,ZK1.remote.cluster.example.org,ZK2.remote.cluster.example.org"
hbase(main):002:0> @shell.hbase.configuration.get("raining")
=> "false"

20. Shell Tricks

20.1. Table variables

HBase 0.95 adds shell commands that provides jruby-style object-oriented references for tables. Previously all of the shell commands that act upon a table have a procedural style that always took the name of the table as an argument. HBase 0.95 introduces the ability to assign a table to a jruby variable. The table reference can be used to perform data read write operations such as puts, scans, and gets well as admin functionality such as disabling, dropping, describing tables.

For example, previously you would always specify a table name:

hbase(main):000:0> create ‘t’, ‘f’
0 row(s) in 1.0970 seconds
hbase(main):001:0> put 't', 'rold', 'f', 'v'
0 row(s) in 0.0080 seconds

hbase(main):002:0> scan 't'
ROW                                COLUMN+CELL
 rold                              column=f:, timestamp=1378473207660, value=v
1 row(s) in 0.0130 seconds

hbase(main):003:0> describe 't'
DESCRIPTION                                                                           ENABLED
 SCOPE => '0', VERSIONS => '1', COMPRESSION => 'NONE', MIN_VERSIONS => '0', TTL => '2
 147483647', KEEP_DELETED_CELLS => 'false', BLOCKSIZE => '65536', IN_MEMORY => 'false
 ', BLOCKCACHE => 'true'}
1 row(s) in 1.4430 seconds

hbase(main):004:0> disable 't'
0 row(s) in 14.8700 seconds

hbase(main):005:0> drop 't'
0 row(s) in 23.1670 seconds


Now you can assign the table to a variable and use the results in jruby shell code.

hbase(main):007 > t = create 't', 'f'
0 row(s) in 1.0970 seconds

=> Hbase::Table - t
hbase(main):008 > t.put 'r', 'f', 'v'
0 row(s) in 0.0640 seconds
hbase(main):009 > t.scan
ROW                           COLUMN+CELL
 r                            column=f:, timestamp=1331865816290, value=v
1 row(s) in 0.0110 seconds
hbase(main):010:0> t.describe
DESCRIPTION                                                                           ENABLED
 SCOPE => '0', VERSIONS => '1', COMPRESSION => 'NONE', MIN_VERSIONS => '0', TTL => '2
 147483647', KEEP_DELETED_CELLS => 'false', BLOCKSIZE => '65536', IN_MEMORY => 'false
 ', BLOCKCACHE => 'true'}
1 row(s) in 0.0210 seconds
hbase(main):038:0> t.disable
0 row(s) in 6.2350 seconds
hbase(main):039:0> t.drop
0 row(s) in 0.2340 seconds

If the table has already been created, you can assign a Table to a variable by using the get_table method:

hbase(main):011 > create 't','f'
0 row(s) in 1.2500 seconds

=> Hbase::Table - t
hbase(main):012:0> tab = get_table 't'
0 row(s) in 0.0010 seconds

=> Hbase::Table - t
hbase(main):013:0> tab.put ‘r1’ ,’f’, ‘v’
0 row(s) in 0.0100 seconds
hbase(main):014:0> tab.scan
ROW                                COLUMN+CELL
 r1                                column=f:, timestamp=1378473876949, value=v
1 row(s) in 0.0240 seconds

The list functionality has also been extended so that it returns a list of table names as strings. You can then use jruby to script table operations based on these names. The list_snapshots command also acts similarly.

hbase(main):016 > tables = list(‘t.*’)
1 row(s) in 0.1040 seconds

=> #<#<Class:0x7677ce29>:0x21d377a4>
hbase(main):017:0> tables.map { |t| disable t ; drop  t}
0 row(s) in 2.2510 seconds

=> [nil]

20.2. irbrc

Create an .irbrc file for yourself in your home directory. Add customizations. A useful one is command history so commands are save across Shell invocations:

$ more .irbrc
require 'irb/ext/save-history'
IRB.conf[:SAVE_HISTORY] = 100
IRB.conf[:HISTORY_FILE] = "#{ENV['HOME']}/.irb-save-history"

See the ruby documentation of .irbrc to learn about other possible configurations.

20.3. LOG data to timestamp

To convert the date '08/08/16 20:56:29' from an hbase log into a timestamp, do:

hbase(main):021:0> import java.text.SimpleDateFormat
hbase(main):022:0> import java.text.ParsePosition
hbase(main):023:0> SimpleDateFormat.new("yy/MM/dd HH:mm:ss").parse("08/08/16 20:56:29", ParsePosition.new(0)).getTime() => 1218920189000

To go the other direction:

hbase(main):021:0> import java.util.Date
hbase(main):022:0> Date.new(1218920189000).toString() => "Sat Aug 16 20:56:29 UTC 2008"

To output in a format that is exactly like that of the HBase log format will take a little messing with SimpleDateFormat.

20.4. Pre-splitting tables with the HBase Shell

You can use a variety of options to pre-split tables when creating them via the HBase Shell create command.

The simplest approach is to specify an array of split points when creating the table. Note that when specifying string literals as split points, these will create split points based on the underlying byte representation of the string. So when specifying a split point of '10', we are actually specifying the byte split point '\x31\30'.

The split points will define n+1 regions where n is the number of split points. The lowest region will contain all keys from the lowest possible key up to but not including the first split point key. The next region will contain keys from the first split point up to, but not including the next split point key. This will continue for all split points up to the last. The last region will be defined from the last split point up to the maximum possible key.

hbase>create 't1','f',SPLITS => ['10','20',30']

In the above example, the table 't1' will be created with column family 'f', pre-split to four regions. Note the first region will contain all keys from '\x00' up to '\x30' (as '\x31' is the ASCII code for '1').

You can pass the split points in a file using following variation. In this example, the splits are read from a file corresponding to the local path on the local filesystem. Each line in the file specifies a split point key.

hbase>create 't14','f',SPLITS_FILE=>'splits.txt'

The other options are to automatically compute splits based on a desired number of regions and a splitting algorithm. HBase supplies algorithms for splitting the key range based on uniform splits or based on hexadecimal keys, but you can provide your own splitting algorithm to subdivide the key range.

# create table with four regions based on random bytes keys
hbase>create 't2','f1', { NUMREGIONS => 4 , SPLITALGO => 'UniformSplit' }

# create table with five regions based on hex keys
hbase>create 't3','f1', { NUMREGIONS => 5, SPLITALGO => 'HexStringSplit' }

As the HBase Shell is effectively a Ruby environment, you can use simple Ruby scripts to compute splits algorithmically.

# generate splits for long (Ruby fixnum) key range from start to end key
hbase(main):070:0> def gen_splits(start_key,end_key,num_regions)
hbase(main):071:1>   results=[]
hbase(main):072:1>   range=end_key-start_key
hbase(main):073:1>   incr=(range/num_regions).floor
hbase(main):074:1>   for i in 1 .. num_regions-1
hbase(main):075:2>     results.push([i*incr+start_key].pack("N"))
hbase(main):076:2>   end
hbase(main):077:1>   return results
hbase(main):078:1> end
hbase(main):080:0> splits=gen_splits(1,2000000,10)
=> ["\000\003\r@", "\000\006\032\177", "\000\t'\276", "\000\f4\375", "\000\017B<", "\000\022O{", "\000\025\\\272", "\000\030i\371", "\000\ew8"]
hbase(main):081:0> create 'test_splits','f',SPLITS=>splits
0 row(s) in 0.2670 seconds

=> Hbase::Table - test_splits

Note that the HBase Shell command truncate effectively drops and recreates the table with default options which will discard any pre-splitting. If you need to truncate a pre-split table, you must drop and recreate the table explicitly to re-specify custom split options.

20.5. Debug

20.5.1. Shell debug switch

You can set a debug switch in the shell to see more output — e.g. more of the stack trace on exception — when you run a command:

hbase> debug <RETURN>

20.5.2. DEBUG log level

To enable DEBUG level logging in the shell, launch it with the -d option.

$ ./bin/hbase shell -d

20.6. Commands

20.6.1. count

Count command returns the number of rows in a table. It’s quite fast when configured with the right CACHE

hbase> count '<tablename>', CACHE => 1000

The above count fetches 1000 rows at a time. Set CACHE lower if your rows are big. Default is to fetch one row at a time.

Data Model

In HBase, data is stored in tables, which have rows and columns. This is a terminology overlap with relational databases (RDBMSs), but this is not a helpful analogy. Instead, it can be helpful to think of an HBase table as a multi-dimensional map.

HBase Data Model Terminology

An HBase table consists of multiple rows.


A row in HBase consists of a row key and one or more columns with values associated with them. Rows are sorted alphabetically by the row key as they are stored. For this reason, the design of the row key is very important. The goal is to store data in such a way that related rows are near each other. A common row key pattern is a website domain. If your row keys are domains, you should probably store them in reverse (org.apache.www, org.apache.mail, org.apache.jira). This way, all of the Apache domains are near each other in the table, rather than being spread out based on the first letter of the subdomain.


A column in HBase consists of a column family and a column qualifier, which are delimited by a : (colon) character.

Column Family

Column families physically colocate a set of columns and their values, often for performance reasons. Each column family has a set of storage properties, such as whether its values should be cached in memory, how its data is compressed or its row keys are encoded, and others. Each row in a table has the same column families, though a given row might not store anything in a given column family.

Column Qualifier

A column qualifier is added to a column family to provide the index for a given piece of data. Given a column family content, a column qualifier might be content:html, and another might be content:pdf. Though column families are fixed at table creation, column qualifiers are mutable and may differ greatly between rows.


A cell is a combination of row, column family, and column qualifier, and contains a value and a timestamp, which represents the value’s version.


A timestamp is written alongside each value, and is the identifier for a given version of a value. By default, the timestamp represents the time on the RegionServer when the data was written, but you can specify a different timestamp value when you put data into the cell.

21. Conceptual View

You can read a very understandable explanation of the HBase data model in the blog post Understanding HBase and BigTable by Jim R. Wilson. Another good explanation is available in the PDF Introduction to Basic Schema Design by Amandeep Khurana.

It may help to read different perspectives to get a solid understanding of HBase schema design. The linked articles cover the same ground as the information in this section.

The following example is a slightly modified form of the one on page 2 of the BigTable paper. There is a table called webtable that contains two rows (com.cnn.www and com.example.www) and three column families named contents, anchor, and people. In this example, for the first row (com.cnn.www), anchor contains two columns (anchor:cssnsi.com, anchor:my.look.ca) and contents contains one column (contents:html). This example contains 5 versions of the row with the row key com.cnn.www, and one version of the row with the row key com.example.www. The contents:html column qualifier contains the entire HTML of a given website. Qualifiers of the anchor column family each contain the external site which links to the site represented by the row, along with the text it used in the anchor of its link. The people column family represents people associated with the site.

Column Names

By convention, a column name is made of its column family prefix and a qualifier. For example, the column contents:html is made up of the column family contents and the html qualifier. The colon character (:) delimits the column family from the column family qualifier.

Table 4. Table webtable
Row Key Time Stamp ColumnFamily contents ColumnFamily anchor ColumnFamily people



anchor:cnnsi.com = "CNN"



anchor:my.look.ca = "CNN.com"



contents:html = "<html>…​"



contents:html = "<html>…​"



contents:html = "<html>…​"

Cells in this table that appear to be empty do not take space, or in fact exist, in HBase. This is what makes HBase "sparse." A tabular view is not the only possible way to look at data in HBase, or even the most accurate. The following represents the same information as a multi-dimensional map. This is only a mock-up for illustrative purposes and may not be strictly accurate.

  "com.cnn.www": {
    contents: {
      t6: contents:html: "<html>..."
      t5: contents:html: "<html>..."
      t3: contents:html: "<html>..."
    anchor: {
      t9: anchor:cnnsi.com = "CNN"
      t8: anchor:my.look.ca = "CNN.com"
    people: {}
  "com.example.www": {
    contents: {
      t5: contents:html: "<html>..."
    anchor: {}
    people: {
      t5: people:author: "John Doe"

22. Physical View

Although at a conceptual level tables may be viewed as a sparse set of rows, they are physically stored by column family. A new column qualifier (column_family:column_qualifier) can be added to an existing column family at any time.

Table 5. ColumnFamily anchor
Row Key Time Stamp Column Family anchor



anchor:cnnsi.com = "CNN"



anchor:my.look.ca = "CNN.com"

Table 6. ColumnFamily contents
Row Key Time Stamp ColumnFamily contents:



contents:html = "<html>…​"



contents:html = "<html>…​"



contents:html = "<html>…​"

The empty cells shown in the conceptual view are not stored at all. Thus a request for the value of the contents:html column at time stamp t8 would return no value. Similarly, a request for an anchor:my.look.ca value at time stamp t9 would return no value. However, if no timestamp is supplied, the most recent value for a particular column would be returned. Given multiple versions, the most recent is also the first one found, since timestamps are stored in descending order. Thus a request for the values of all columns in the row com.cnn.www if no timestamp is specified would be: the value of contents:html from timestamp t6, the value of anchor:cnnsi.com from timestamp t9, the value of anchor:my.look.ca from timestamp t8.

For more information about the internals of how Apache HBase stores data, see regions.arch.

23. Namespace

A namespace is a logical grouping of tables analogous to a database in relation database systems. This abstraction lays the groundwork for upcoming multi-tenancy related features:

  • Quota Management (HBASE-8410) - Restrict the amount of resources (i.e. regions, tables) a namespace can consume.

  • Namespace Security Administration (HBASE-9206) - Provide another level of security administration for tenants.

  • Region server groups (HBASE-6721) - A namespace/table can be pinned onto a subset of RegionServers thus guaranteeing a course level of isolation.

23.1. Namespace management

A namespace can be created, removed or altered. Namespace membership is determined during table creation by specifying a fully-qualified table name of the form:

<table namespace>:<table qualifier>
Example 12. Examples
#Create a namespace
create_namespace 'my_ns'
#create my_table in my_ns namespace
create 'my_ns:my_table', 'fam'
#drop namespace
drop_namespace 'my_ns'
#alter namespace
alter_namespace 'my_ns', {METHOD => 'set', 'PROPERTY_NAME' => 'PROPERTY_VALUE'}

23.2. Predefined namespaces

There are two predefined special namespaces:

  • hbase - system namespace, used to contain HBase internal tables

  • default - tables with no explicit specified namespace will automatically fall into this namespace

Example 13. Examples
#namespace=foo and table qualifier=bar
create 'foo:bar', 'fam'

#namespace=default and table qualifier=bar
create 'bar', 'fam'

24. Table

Tables are declared up front at schema definition time.

25. Row

Row keys are uninterpreted bytes. Rows are lexicographically sorted with the lowest order appearing first in a table. The empty byte array is used to denote both the start and end of a tables' namespace.

26. Column Family

Columns in Apache HBase are grouped into column families. All column members of a column family have the same prefix. For example, the columns courses:history and courses:math are both members of the courses column family. The colon character (:) delimits the column family from the column family qualifier. The column family prefix must be composed of printable characters. The qualifying tail, the column family qualifier, can be made of any arbitrary bytes. Column families must be declared up front at schema definition time whereas columns do not need to be defined at schema time but can be conjured on the fly while the table is up and running.

Physically, all column family members are stored together on the filesystem. Because tunings and storage specifications are done at the column family level, it is advised that all column family members have the same general access pattern and size characteristics.

27. Cells

A {row, column, version} tuple exactly specifies a cell in HBase. Cell content is uninterpreted bytes

28. Data Model Operations

The four primary data model operations are Get, Put, Scan, and Delete. Operations are applied via Table instances.

28.1. Get

Get returns attributes for a specified row. Gets are executed via Table.get.

28.2. Put

Put either adds new rows to a table (if the key is new) or can update existing rows (if the key already exists). Puts are executed via Table.put (writeBuffer) or Table.batch (non-writeBuffer).

28.3. Scans

Scan allow iteration over multiple rows for specified attributes.

The following is an example of a Scan on a Table instance. Assume that a table is populated with rows with keys "row1", "row2", "row3", and then another set of rows with the keys "abc1", "abc2", and "abc3". The following example shows how to set a Scan instance to return the rows beginning with "row".

public static final byte[] CF = "cf".getBytes();
public static final byte[] ATTR = "attr".getBytes();

Table table = ...      // instantiate a Table instance

Scan scan = new Scan();
scan.addColumn(CF, ATTR);
ResultScanner rs = table.getScanner(scan);
try {
  for (Result r = rs.next(); r != null; r = rs.next()) {
    // process result...
} finally {
  rs.close();  // always close the ResultScanner!

Note that generally the easiest way to specify a specific stop point for a scan is by using the InclusiveStopFilter class.

28.4. Delete

Delete removes a row from a table. Deletes are executed via Table.delete.

HBase does not modify data in place, and so deletes are handled by creating new markers called tombstones. These tombstones, along with the dead values, are cleaned up on major compactions.

See version.delete for more information on deleting versions of columns, and see compaction for more information on compactions.

29. Versions

A {row, column, version} tuple exactly specifies a cell in HBase. It’s possible to have an unbounded number of cells where the row and column are the same but the cell address differs only in its version dimension.

While rows and column keys are expressed as bytes, the version is specified using a long integer. Typically this long contains time instances such as those returned by java.util.Date.getTime() or System.currentTimeMillis(), that is: the difference, measured in milliseconds, between the current time and midnight, January 1, 1970 UTC.

The HBase version dimension is stored in decreasing order, so that when reading from a store file, the most recent values are found first.

There is a lot of confusion over the semantics of cell versions, in HBase. In particular:

  • If multiple writes to a cell have the same version, only the last written is fetchable.

  • It is OK to write cells in a non-increasing version order.

Below we describe how the version dimension in HBase currently works. See HBASE-2406 for discussion of HBase versions. Bending time in HBase makes for a good read on the version, or time, dimension in HBase. It has more detail on versioning than is provided here. As of this writing, the limitation Overwriting values at existing timestamps mentioned in the article no longer holds in HBase. This section is basically a synopsis of this article by Bruno Dumon.

29.1. Specifying the Number of Versions to Store

The maximum number of versions to store for a given column is part of the column schema and is specified at table creation, or via an alter command, via HColumnDescriptor.DEFAULT_VERSIONS. Prior to HBase 0.96, the default number of versions kept was 3, but in 0.96 and newer has been changed to 1.

Example 14. Modify the Maximum Number of Versions for a Column Family

This example uses HBase Shell to keep a maximum of 5 versions of all columns in column family f1. You could also use HColumnDescriptor.

hbase> alter ‘t1′, NAME => ‘f1′, VERSIONS => 5
Example 15. Modify the Minimum Number of Versions for a Column Family

You can also specify the minimum number of versions to store per column family. By default, this is set to 0, which means the feature is disabled. The following example sets the minimum number of versions on all columns in column family f1 to 2, via HBase Shell. You could also use HColumnDescriptor.

hbase> alter ‘t1′, NAME => ‘f1′, MIN_VERSIONS => 2

Starting with HBase 0.98.2, you can specify a global default for the maximum number of versions kept for all newly-created columns, by setting hbase.column.max.version in hbase-site.xml. See hbase.column.max.version.

29.2. Versions and HBase Operations

In this section we look at the behavior of the version dimension for each of the core HBase operations.

29.2.1. Get/Scan

Gets are implemented on top of Scans. The below discussion of Get applies equally to Scans.

By default, i.e. if you specify no explicit version, when doing a get, the cell whose version has the largest value is returned (which may or may not be the latest one written, see later). The default behavior can be modified in the following ways:

  • to return more than one version, see Get.setMaxVersions()

  • to return versions other than the latest, see Get.setTimeRange()

    To retrieve the latest version that is less than or equal to a given value, thus giving the 'latest' state of the record at a certain point in time, just use a range from 0 to the desired version and set the max versions to 1.

29.2.2. Default Get Example

The following Get will only retrieve the current version of the row

public static final byte[] CF = "cf".getBytes();
public static final byte[] ATTR = "attr".getBytes();
Get get = new Get(Bytes.toBytes("row1"));
Result r = table.get(get);
byte[] b = r.getValue(CF, ATTR);  // returns current version of value

29.2.3. Versioned Get Example

The following Get will return the last 3 versions of the row.

public static final byte[] CF = "cf".getBytes();
public static final byte[] ATTR = "attr".getBytes();
Get get = new Get(Bytes.toBytes("row1"));
get.setMaxVersions(3);  // will return last 3 versions of row
Result r = table.get(get);
byte[] b = r.getValue(CF, ATTR);  // returns current version of value
List<KeyValue> kv = r.getColumn(CF, ATTR);  // returns all versions of this column

29.2.4. Put

Doing a put always creates a new version of a cell, at a certain timestamp. By default the system uses the server’s currentTimeMillis, but you can specify the version (= the long integer) yourself, on a per-column level. This means you could assign a time in the past or the future, or use the long value for non-time purposes.

To overwrite an existing value, do a put at exactly the same row, column, and version as that of the cell you want to overwrite.

Implicit Version Example

The following Put will be implicitly versioned by HBase with the current time.

public static final byte[] CF = "cf".getBytes();
public static final byte[] ATTR = "attr".getBytes();
Put put = new Put(Bytes.toBytes(row));
put.add(CF, ATTR, Bytes.toBytes( data));
Explicit Version Example

The following Put has the version timestamp explicitly set.

public static final byte[] CF = "cf".getBytes();
public static final byte[] ATTR = "attr".getBytes();
Put put = new Put( Bytes.toBytes(row));
long explicitTimeInMs = 555;  // just an example
put.add(CF, ATTR, explicitTimeInMs, Bytes.toBytes(data));

Caution: the version timestamp is used internally by HBase for things like time-to-live calculations. It’s usually best to avoid setting this timestamp yourself. Prefer using a separate timestamp attribute of the row, or have the timestamp as a part of the row key, or both.

29.2.5. Delete

There are three different types of internal delete markers. See Lars Hofhansl’s blog for discussion of his attempt adding another, Scanning in HBase: Prefix Delete Marker.

  • Delete: for a specific version of a column.

  • Delete column: for all versions of a column.

  • Delete family: for all columns of a particular ColumnFamily

When deleting an entire row, HBase will internally create a tombstone for each ColumnFamily (i.e., not each individual column).

Deletes work by creating tombstone markers. For example, let’s suppose we want to delete a row. For this you can specify a version, or else by default the currentTimeMillis is used. What this means is delete all cells where the version is less than or equal to this version. HBase never modifies data in place, so for example a delete will not immediately delete (or mark as deleted) the entries in the storage file that correspond to the delete condition. Rather, a so-called tombstone is written, which will mask the deleted values. When HBase does a major compaction, the tombstones are processed to actually remove the dead values, together with the tombstones themselves. If the version you specified when deleting a row is larger than the version of any value in the row, then you can consider the complete row to be deleted.

For an informative discussion on how deletes and versioning interact, see the thread Put w/timestamp → Deleteall → Put w/ timestamp fails up on the user mailing list.

Also see keyvalue for more information on the internal KeyValue format.

Delete markers are purged during the next major compaction of the store, unless the KEEP_DELETED_CELLS option is set in the column family (See Keeping Deleted Cells). To keep the deletes for a configurable amount of time, you can set the delete TTL via the hbase.hstore.time.to.purge.deletes property in hbase-site.xml. If hbase.hstore.time.to.purge.deletes is not set, or set to 0, all delete markers, including those with timestamps in the future, are purged during the next major compaction. Otherwise, a delete marker with a timestamp in the future is kept until the major compaction which occurs after the time represented by the marker’s timestamp plus the value of hbase.hstore.time.to.purge.deletes, in milliseconds.

This behavior represents a fix for an unexpected change that was introduced in HBase 0.94, and was fixed in HBASE-10118. The change has been backported to HBase 0.94 and newer branches.

29.3. Current Limitations

29.3.1. Deletes mask Puts

Deletes mask puts, even puts that happened after the delete was entered. See HBASE-2256. Remember that a delete writes a tombstone, which only disappears after then next major compaction has run. Suppose you do a delete of everything ⇐ T. After this you do a new put with a timestamp ⇐ T. This put, even if it happened after the delete, will be masked by the delete tombstone. Performing the put will not fail, but when you do a get you will notice the put did have no effect. It will start working again after the major compaction has run. These issues should not be a problem if you use always-increasing versions for new puts to a row. But they can occur even if you do not care about time: just do delete and put immediately after each other, and there is some chance they happen within the same millisecond.

29.3.2. Major compactions change query results

…​create three cell versions at t1, t2 and t3, with a maximum-versions setting of 2. So when getting all versions, only the values at t2 and t3 will be returned. But if you delete the version at t2 or t3, the one at t1 will appear again. Obviously, once a major compaction has run, such behavior will not be the case anymore…​ (See Garbage Collection in Bending time in HBase.)

30. Sort Order

All data model operations HBase return data in sorted order. First by row, then by ColumnFamily, followed by column qualifier, and finally timestamp (sorted in reverse, so newest records are returned first).

31. Column Metadata

There is no store of column metadata outside of the internal KeyValue instances for a ColumnFamily. Thus, while HBase can support not only a wide number of columns per row, but a heterogeneous set of columns between rows as well, it is your responsibility to keep track of the column names.

The only way to get a complete set of columns that exist for a ColumnFamily is to process all the rows. For more information about how HBase stores data internally, see keyvalue.

32. Joins

Whether HBase supports joins is a common question on the dist-list, and there is a simple answer: it doesn’t, at not least in the way that RDBMS' support them (e.g., with equi-joins or outer-joins in SQL). As has been illustrated in this chapter, the read data model operations in HBase are Get and Scan.

However, that doesn’t mean that equivalent join functionality can’t be supported in your application, but you have to do it yourself. The two primary strategies are either denormalizing the data upon writing to HBase, or to have lookup tables and do the join between HBase tables in your application or MapReduce code (and as RDBMS' demonstrate, there are several strategies for this depending on the size of the tables, e.g., nested loops vs. hash-joins). So which is the best approach? It depends on what you are trying to do, and as such there isn’t a single answer that works for every use case.

33. ACID

See ACID Semantics. Lars Hofhansl has also written a note on ACID in HBase.

HBase and Schema Design

A good introduction on the strength and weaknesses modelling on the various non-rdbms datastores is to be found in Ian Varley’s Master thesis, No Relation: The Mixed Blessings of Non-Relational Databases. It is a little dated now but a good background read if you have a moment on how HBase schema modeling differs from how it is done in an RDBMS. Also, read keyvalue for how HBase stores data internally, and the section on schema.casestudies.

The documentation on the Cloud Bigtable website, Designing Your Schema, is pertinent and nicely done and lessons learned there equally apply here in HBase land; just divide any quoted values by ~10 to get what works for HBase: e.g. where it says individual values can be ~10MBs in size, HBase can do similar — perhaps best to go smaller if you can — and where it says a maximum of 100 column families in Cloud Bigtable, think ~10 when modeling on HBase.

34. Schema Creation

HBase schemas can be created or updated using the The Apache HBase Shell or by using Admin in the Java API.

Tables must be disabled when making ColumnFamily modifications, for example:

Configuration config = HBaseConfiguration.create();
Admin admin = new Admin(conf);
TableName table = TableName.valueOf("myTable");


HColumnDescriptor cf1 = ...;
admin.addColumn(table, cf1);      // adding new ColumnFamily
HColumnDescriptor cf2 = ...;
admin.modifyColumn(table, cf2);    // modifying existing ColumnFamily


See client dependencies for more information about configuring client connections.

online schema changes are supported in the 0.92.x codebase, but the 0.90.x codebase requires the table to be disabled.

34.1. Schema Updates

When changes are made to either Tables or ColumnFamilies (e.g. region size, block size), these changes take effect the next time there is a major compaction and the StoreFiles get re-written.

See store for more information on StoreFiles.

35. Table Schema Rules Of Thumb

There are many different data sets, with different access patterns and service-level expectations. Therefore, these rules of thumb are only an overview. Read the rest of this chapter to get more details after you have gone through this list.

  • Aim to have regions sized between 10 and 50 GB.

  • Aim to have cells no larger than 10 MB, or 50 MB if you use mob. Otherwise, consider storing your cell data in HDFS and store a pointer to the data in HBase.

  • A typical schema has between 1 and 3 column families per table. HBase tables should not be designed to mimic RDBMS tables.

  • Around 50-100 regions is a good number for a table with 1 or 2 column families. Remember that a region is a contiguous segment of a column family.

  • Keep your column family names as short as possible. The column family names are stored for every value (ignoring prefix encoding). They should not be self-documenting and descriptive like in a typical RDBMS.

  • If you are storing time-based machine data or logging information, and the row key is based on device ID or service ID plus time, you can end up with a pattern where older data regions never have additional writes beyond a certain age. In this type of situation, you end up with a small number of active regions and a large number of older regions which have no new writes. For these situations, you can tolerate a larger number of regions because your resource consumption is driven by the active regions only.

  • If only one column family is busy with writes, only that column family accomulates memory. Be aware of write patterns when allocating resources.

RegionServer Sizing Rules of Thumb

Lars Hofhansl wrote a great blog post about RegionServer memory sizing. The upshot is that you probably need more memory than you think you need. He goes into the impact of region size, memstore size, HDFS replication factor, and other things to check.

Personally I would place the maximum disk space per machine that can be served exclusively with HBase around 6T, unless you have a very read-heavy workload. In that case the Java heap should be 32GB (20G regions, 128M memstores, the rest defaults).

— Lars Hofhansl

36. On the number of column families

HBase currently does not do well with anything above two or three column families so keep the number of column families in your schema low. Currently, flushing and compactions are done on a per Region basis so if one column family is carrying the bulk of the data bringing on flushes, the adjacent families will also be flushed even though the amount of data they carry is small. When many column families exist the flushing and compaction interaction can make for a bunch of needless i/o (To be addressed by changing flushing and compaction to work on a per column family basis). For more information on compactions, see Compaction.

Try to make do with one column family if you can in your schemas. Only introduce a second and third column family in the case where data access is usually column scoped; i.e. you query one column family or the other but usually not both at the one time.

36.1. Cardinality of ColumnFamilies

Where multiple ColumnFamilies exist in a single table, be aware of the cardinality (i.e., number of rows). If ColumnFamilyA has 1 million rows and ColumnFamilyB has 1 billion rows, ColumnFamilyA’s data will likely be spread across many, many regions (and RegionServers). This makes mass scans for ColumnFamilyA less efficient.

37. Rowkey Design

37.1. Hotspotting

Rows in HBase are sorted lexicographically by row key. This design optimizes for scans, allowing you to store related rows, or rows that will be read together, near each other. However, poorly designed row keys are a common source of hotspotting. Hotspotting occurs when a large amount of client traffic is directed at one node, or only a few nodes, of a cluster. This traffic may represent reads, writes, or other operations. The traffic overwhelms the single machine responsible for hosting that region, causing performance degradation and potentially leading to region unavailability. This can also have adverse effects on other regions hosted by the same region server as that host is unable to service the requested load. It is important to design data access patterns such that the cluster is fully and evenly utilized.

To prevent hotspotting on writes, design your row keys such that rows that truly do need to be in the same region are, but in the bigger picture, data is being written to multiple regions across the cluster, rather than one at a time. Some common techniques for avoiding hotspotting are described below, along with some of their advantages and drawbacks.


Salting in this sense has nothing to do with cryptography, but refers to adding random data to the start of a row key. In this case, salting refers to adding a randomly-assigned prefix to the row key to cause it to sort differently than it otherwise would. The number of possible prefixes correspond to the number of regions you want to spread the data across. Salting can be helpful if you have a few "hot" row key patterns which come up over and over amongst other more evenly-distributed rows. Consider the following example, which shows that salting can spread write load across multiple RegionServers, and illustrates some of the negative implications for reads.

Example 16. Salting Example

Suppose you have the following list of row keys, and your table is split such that there is one region for each letter of the alphabet. Prefix 'a' is one region, prefix 'b' is another. In this table, all rows starting with 'f' are in the same region. This example focuses on rows with keys like the following:


Now, imagine that you would like to spread these across four different regions. You decide to use four different salts: a, b, c, and d. In this scenario, each of these letter prefixes will be on a different region. After applying the salts, you have the following rowkeys instead. Since you can now write to four separate regions, you theoretically have four times the throughput when writing that you would have if all the writes were going to the same region.


Then, if you add another row, it will randomly be assigned one of the four possible salt values and end up near one of the existing rows.


Since this assignment will be random, you will need to do more work if you want to retrieve the rows in lexicographic order. In this way, salting attempts to increase throughput on writes, but has a cost during reads.


Instead of a random assignment, you could use a one-way hash that would cause a given row to always be "salted" with the same prefix, in a way that would spread the load across the RegionServers, but allow for predictability during reads. Using a deterministic hash allows the client to reconstruct the complete rowkey and use a Get operation to retrieve that row as normal.

Example 17. Hashing Example
Given the same situation in the salting example above, you could instead apply a one-way hash that would cause the row with key foo0003 to always, and predictably, receive the a prefix. Then, to retrieve that row, you would already know the key. You could also optimize things so that certain pairs of keys were always in the same region, for instance.
Reversing the Key

A third common trick for preventing hotspotting is to reverse a fixed-width or numeric row key so that the part that changes the most often (the least significant digit) is first. This effectively randomizes row keys, but sacrifices row ordering properties.

See https://communities.intel.com/community/itpeernetwork/datastack/blog/2013/11/10/discussion-on-designing-hbase-tables, and article on Salted Tables from the Phoenix project, and the discussion in the comments of HBASE-11682 for more information about avoiding hotspotting.

37.2. Monotonically Increasing Row Keys/Timeseries Data

In the HBase chapter of Tom White’s book Hadoop: The Definitive Guide (O’Reilly) there is a an optimization note on watching out for a phenomenon where an import process walks in lock-step with all clients in concert pounding one of the table’s regions (and thus, a single node), then moving onto the next region, etc. With monotonically increasing row-keys (i.e., using a timestamp), this will happen. See this comic by IKai Lan on why monotonically increasing row keys are problematic in BigTable-like datastores: monotonically increasing values are bad. The pile-up on a single region brought on by monotonically increasing keys can be mitigated by randomizing the input records to not be in sorted order, but in general it’s best to avoid using a timestamp or a sequence (e.g. 1, 2, 3) as the row-key.

If you do need to upload time series data into HBase, you should study OpenTSDB as a successful example. It has a page describing the schema it uses in HBase. The key format in OpenTSDB is effectively [metric_type][event_timestamp], which would appear at first glance to contradict the previous advice about not using a timestamp as the key. However, the difference is that the timestamp is not in the lead position of the key, and the design assumption is that there are dozens or hundreds (or more) of different metric types. Thus, even with a continual stream of input data with a mix of metric types, the Puts are distributed across various points of regions in the table.

See schema.casestudies for some rowkey design examples.

37.3. Try to minimize row and column sizes

In HBase, values are always freighted with their coordinates; as a cell value passes through the system, it’ll be accompanied by its row, column name, and timestamp - always. If your rows and column names are large, especially compared to the size of the cell value, then you may run up against some interesting scenarios. One such is the case described by Marc Limotte at the tail of HBASE-3551 (recommended!). Therein, the indices that are kept on HBase storefiles (StoreFile (HFile)) to facilitate random access may end up occupying large chunks of the HBase allotted RAM because the cell value coordinates are large. Mark in the above cited comment suggests upping the block size so entries in the store file index happen at a larger interval or modify the table schema so it makes for smaller rows and column names. Compression will also make for larger indices. See the thread a question storefileIndexSize up on the user mailing list.

Most of the time small inefficiencies don’t matter all that much. Unfortunately, this is a case where they do. Whatever patterns are selected for ColumnFamilies, attributes, and rowkeys they could be repeated several billion times in your data.

See keyvalue for more information on HBase stores data internally to see why this is important.

37.3.1. Column Families

Try to keep the ColumnFamily names as small as possible, preferably one character (e.g. "d" for data/default).

See KeyValue for more information on HBase stores data internally to see why this is important.

37.3.2. Attributes

Although verbose attribute names (e.g., "myVeryImportantAttribute") are easier to read, prefer shorter attribute names (e.g., "via") to store in HBase.

See keyvalue for more information on HBase stores data internally to see why this is important.

37.3.3. Rowkey Length

Keep them as short as is reasonable such that they can still be useful for required data access (e.g. Get vs. Scan). A short key that is useless for data access is not better than a longer key with better get/scan properties. Expect tradeoffs when designing rowkeys.

37.3.4. Byte Patterns

A long is 8 bytes. You can store an unsigned number up to 18,446,744,073,709,551,615 in those eight bytes. If you stored this number as a String — presuming a byte per character — you need nearly 3x the bytes.

Not convinced? Below is some sample code that you can run on your own.

// long
long l = 1234567890L;
byte[] lb = Bytes.toBytes(l);
System.out.println("long bytes length: " + lb.length);   // returns 8

String s = String.valueOf(l);
byte[] sb = Bytes.toBytes(s);
System.out.println("long as string length: " + sb.length);    // returns 10

// hash
MessageDigest md = MessageDigest.getInstance("MD5");
byte[] digest = md.digest(Bytes.toBytes(s));
System.out.println("md5 digest bytes length: " + digest.length);    // returns 16

String sDigest = new String(digest);
byte[] sbDigest = Bytes.toBytes(sDigest);
System.out.println("md5 digest as string length: " + sbDigest.length);    // returns 26

Unfortunately, using a binary representation of a type will make your data harder to read outside of your code. For example, this is what you will see in the shell when you increment a value:

hbase(main):001:0> incr 't', 'r', 'f:q', 1

hbase(main):002:0> get 't', 'r'
COLUMN                                        CELL
 f:q                                          timestamp=1369163040570, value=\x00\x00\x00\x00\x00\x00\x00\x01
1 row(s) in 0.0310 seconds

The shell makes a best effort to print a string, and it this case it decided to just print the hex. The same will happen to your row keys inside the region names. It can be okay if you know what’s being stored, but it might also be unreadable if arbitrary data can be put in the same cells. This is the main trade-off.

37.4. Reverse Timestamps

Reverse Scan API

HBASE-4811 implements an API to scan a table or a range within a table in reverse, reducing the need to optimize your schema for forward or reverse scanning. This feature is available in HBase 0.98 and later. See https://hbase.apache.org/apidocs/org/apache/hadoop/hbase/client/Scan.html#setReversed%28boolean for more information.

A common problem in database processing is quickly finding the most recent version of a value. A technique using reverse timestamps as a part of the key can help greatly with a special case of this problem. Also found in the HBase chapter of Tom White’s book Hadoop: The Definitive Guide (O’Reilly), the technique involves appending (Long.MAX_VALUE - timestamp) to the end of any key, e.g. [key][reverse_timestamp].

The most recent value for [key] in a table can be found by performing a Scan for [key] and obtaining the first record. Since HBase keys are in sorted order, this key sorts before any older row-keys for [key] and thus is first.

This technique would be used instead of using Number of Versions where the intent is to hold onto all versions "forever" (or a very long time) and at the same time quickly obtain access to any other version by using the same Scan technique.

37.5. Rowkeys and ColumnFamilies

Rowkeys are scoped to ColumnFamilies. Thus, the same rowkey could exist in each ColumnFamily that exists in a table without collision.

37.6. Immutability of Rowkeys

Rowkeys cannot be changed. The only way they can be "changed" in a table is if the row is deleted and then re-inserted. This is a fairly common question on the HBase dist-list so it pays to get the rowkeys right the first time (and/or before you’ve inserted a lot of data).

37.7. Relationship Between RowKeys and Region Splits

If you pre-split your table, it is critical to understand how your rowkey will be distributed across the region boundaries. As an example of why this is important, consider the example of using displayable hex characters as the lead position of the key (e.g., "0000000000000000" to "ffffffffffffffff"). Running those key ranges through Bytes.split (which is the split strategy used when creating regions in Admin.createTable(byte[] startKey, byte[] endKey, numRegions) for 10 regions will generate the following splits…​

48 48 48 48 48 48 48 48 48 48 48 48 48 48 48 48                                // 0
54 -10 -10 -10 -10 -10 -10 -10 -10 -10 -10 -10 -10 -10 -10 -10                 // 6
61 -67 -67 -67 -67 -67 -67 -67 -67 -67 -67 -67 -67 -67 -67 -68                 // =
68 -124 -124 -124 -124 -124 -124 -124 -124 -124 -124 -124 -124 -124 -124 -126  // D
75 75 75 75 75 75 75 75 75 75 75 75 75 75 75 72                                // K
82 18 18 18 18 18 18 18 18 18 18 18 18 18 18 14                                // R
88 -40 -40 -40 -40 -40 -40 -40 -40 -40 -40 -40 -40 -40 -40 -44                 // X
95 -97 -97 -97 -97 -97 -97 -97 -97 -97 -97 -97 -97 -97 -97 -102                // _
102 102 102 102 102 102 102 102 102 102 102 102 102 102 102 102                // f

(note: the lead byte is listed to the right as a comment.) Given that the first split is a '0' and the last split is an 'f', everything is great, right? Not so fast.

The problem is that all the data is going to pile up in the first 2 regions and the last region thus creating a "lumpy" (and possibly "hot") region problem. To understand why, refer to an ASCII Table. '0' is byte 48, and 'f' is byte 102, but there is a huge gap in byte values (bytes 58 to 96) that will never appear in this keyspace because the only values are [0-9] and [a-f]. Thus, the middle regions will never be used. To make pre-splitting work with this example keyspace, a custom definition of splits (i.e., and not relying on the built-in split method) is required.

Lesson #1: Pre-splitting tables is generally a best practice, but you need to pre-split them in such a way that all the regions are accessible in the keyspace. While this example demonstrated the problem with a hex-key keyspace, the same problem can happen with any keyspace. Know your data.

Lesson #2: While generally not advisable, using hex-keys (and more generally, displayable data) can still work with pre-split tables as long as all the created regions are accessible in the keyspace.

To conclude this example, the following is an example of how appropriate splits can be pre-created for hex-keys:.

public static boolean createTable(Admin admin, HTableDescriptor table, byte[][] splits)
throws IOException {
  try {
    admin.createTable( table, splits );
    return true;
  } catch (TableExistsException e) {
    logger.info("table " + table.getNameAsString() + " already exists");
    // the table already exists...
    return false;

public static byte[][] getHexSplits(String startKey, String endKey, int numRegions) {
  byte[][] splits = new byte[numRegions-1][];
  BigInteger lowestKey = new BigInteger(startKey, 16);
  BigInteger highestKey = new BigInteger(endKey, 16);
  BigInteger range = highestKey.subtract(lowestKey);
  BigInteger regionIncrement = range.divide(BigInteger.valueOf(numRegions));
  lowestKey = lowestKey.add(regionIncrement);
  for(int i=0; i < numRegions-1;i++) {
    BigInteger key = lowestKey.add(regionIncrement.multiply(BigInteger.valueOf(i)));
    byte[] b = String.format("%016x", key).getBytes();
    splits[i] = b;
  return splits;

38. Number of Versions

38.1. Maximum Number of Versions

The maximum number of row versions to store is configured per column family via HColumnDescriptor. The default for max versions is 1. This is an important parameter because as described in Data Model section HBase does not overwrite row values, but rather stores different values per row by time (and qualifier). Excess versions are removed during major compactions. The number of max versions may need to be increased or decreased depending on application needs.

It is not recommended setting the number of max versions to an exceedingly high level (e.g., hundreds or more) unless those old values are very dear to you because this will greatly increase StoreFile size.

38.2. Minimum Number of Versions

Like maximum number of row versions, the minimum number of row versions to keep is configured per column family via HColumnDescriptor. The default for min versions is 0, which means the feature is disabled. The minimum number of row versions parameter is used together with the time-to-live parameter and can be combined with the number of row versions parameter to allow configurations such as "keep the last T minutes worth of data, at most N versions, but keep at least M versions around" (where M is the value for minimum number of row versions, M<N). This parameter should only be set when time-to-live is enabled for a column family and must be less than the number of row versions.

39. Supported Datatypes

HBase supports a "bytes-in/bytes-out" interface via Put and Result, so anything that can be converted to an array of bytes can be stored as a value. Input could be strings, numbers, complex objects, or even images as long as they can rendered as bytes.

There are practical limits to the size of values (e.g., storing 10-50MB objects in HBase would probably be too much to ask); search the mailing list for conversations on this topic. All rows in HBase conform to the Data Model, and that includes versioning. Take that into consideration when making your design, as well as block size for the ColumnFamily.

39.1. Counters

One supported datatype that deserves special mention are "counters" (i.e., the ability to do atomic increments of numbers). See Increment in Table.

Synchronization on counters are done on the RegionServer, not in the client.

40. Joins

If you have multiple tables, don’t forget to factor in the potential for Joins into the schema design.

41. Time To Live (TTL)

ColumnFamilies can set a TTL length in seconds, and HBase will automatically delete rows once the expiration time is reached. This applies to all versions of a row - even the current one. The TTL time encoded in the HBase for the row is specified in UTC.

Store files which contains only expired rows are deleted on minor compaction. Setting hbase.store.delete.expired.storefile to false disables this feature. Setting minimum number of versions to other than 0 also disables this.

See HColumnDescriptor for more information.

Recent versions of HBase also support setting time to live on a per cell basis. See HBASE-10560 for more information. Cell TTLs are submitted as an attribute on mutation requests (Appends, Increments, Puts, etc.) using Mutation#setTTL. If the TTL attribute is set, it will be applied to all cells updated on the server by the operation. There are two notable differences between cell TTL handling and ColumnFamily TTLs:

  • Cell TTLs are expressed in units of milliseconds instead of seconds.

  • A cell TTLs cannot extend the effective lifetime of a cell beyond a ColumnFamily level TTL setting.

42. Keeping Deleted Cells

By default, delete markers extend back to the beginning of time. Therefore, Get or Scan operations will not see a deleted cell (row or column), even when the Get or Scan operation indicates a time range before the delete marker was placed.

ColumnFamilies can optionally keep deleted cells. In this case, deleted cells can still be retrieved, as long as these operations specify a time range that ends before the timestamp of any delete that would affect the cells. This allows for point-in-time queries even in the presence of deletes.

Deleted cells are still subject to TTL and there will never be more than "maximum number of versions" deleted cells. A new "raw" scan options returns all deleted rows and the delete markers.

Example 18. Change the Value of KEEP_DELETED_CELLS Using HBase Shell
hbase> hbase> alter ‘t1′, NAME => ‘f1′, KEEP_DELETED_CELLS => true
Example 19. Change the Value of KEEP_DELETED_CELLS Using the API

Let us illustrate the basic effect of setting the KEEP_DELETED_CELLS attribute on a table.

First, without:

create 'test', {NAME=>'e', VERSIONS=>2147483647}
put 'test', 'r1', 'e:c1', 'value', 10
put 'test', 'r1', 'e:c1', 'value', 12
put 'test', 'r1', 'e:c1', 'value', 14
delete 'test', 'r1', 'e:c1',  11

hbase(main):017:0> scan 'test', {RAW=>true, VERSIONS=>1000}
ROW                                              COLUMN+CELL
 r1                                              column=e:c1, timestamp=14, value=value
 r1                                              column=e:c1, timestamp=12, value=value
 r1                                              column=e:c1, timestamp=11, type=DeleteColumn
 r1                                              column=e:c1, timestamp=10, value=value
1 row(s) in 0.0120 seconds

hbase(main):018:0> flush 'test'
0 row(s) in 0.0350 seconds

hbase(main):019:0> scan 'test', {RAW=>true, VERSIONS=>1000}
ROW                                              COLUMN+CELL
 r1                                              column=e:c1, timestamp=14, value=value
 r1                                              column=e:c1, timestamp=12, value=value
 r1                                              column=e:c1, timestamp=11, type=DeleteColumn
1 row(s) in 0.0120 seconds

hbase(main):020:0> major_compact 'test'
0 row(s) in 0.0260 seconds

hbase(main):021:0> scan 'test', {RAW=>true, VERSIONS=>1000}
ROW                                              COLUMN+CELL
 r1                                              column=e:c1, timestamp=14, value=value
 r1                                              column=e:c1, timestamp=12, value=value
1 row(s) in 0.0120 seconds

Notice how delete cells are let go.

Now let’s run the same test only with KEEP_DELETED_CELLS set on the table (you can do table or per-column-family):

hbase(main):005:0> create 'test', {NAME=>'e', VERSIONS=>2147483647, KEEP_DELETED_CELLS => true}
0 row(s) in 0.2160 seconds

=> Hbase::Table - test
hbase(main):006:0> put 'test', 'r1', 'e:c1', 'value', 10
0 row(s) in 0.1070 seconds

hbase(main):007:0> put 'test', 'r1', 'e:c1', 'value', 12
0 row(s) in 0.0140 seconds

hbase(main):008:0> put 'test', 'r1', 'e:c1', 'value', 14
0 row(s) in 0.0160 seconds

hbase(main):009:0> delete 'test', 'r1', 'e:c1',  11
0 row(s) in 0.0290 seconds

hbase(main):010:0> scan 'test', {RAW=>true, VERSIONS=>1000}
ROW                                                                                          COLUMN+CELL
 r1                                                                                          column=e:c1, timestamp=14, value=value
 r1                                                                                          column=e:c1, timestamp=12, value=value
 r1                                                                                          column=e:c1, timestamp=11, type=DeleteColumn
 r1                                                                                          column=e:c1, timestamp=10, value=value
1 row(s) in 0.0550 seconds

hbase(main):011:0> flush 'test'
0 row(s) in 0.2780 seconds

hbase(main):012:0> scan 'test', {RAW=>true, VERSIONS=>1000}
ROW                                                                                          COLUMN+CELL
 r1                                                                                          column=e:c1, timestamp=14, value=value
 r1                                                                                          column=e:c1, timestamp=12, value=value
 r1                                                                                          column=e:c1, timestamp=11, type=DeleteColumn
 r1                                                                                          column=e:c1, timestamp=10, value=value
1 row(s) in 0.0620 seconds

hbase(main):013:0> major_compact 'test'
0 row(s) in 0.0530 seconds

hbase(main):014:0> scan 'test', {RAW=>true, VERSIONS=>1000}
ROW                                                                                          COLUMN+CELL
 r1                                                                                          column=e:c1, timestamp=14, value=value
 r1                                                                                          column=e:c1, timestamp=12, value=value
 r1                                                                                          column=e:c1, timestamp=11, type=DeleteColumn
 r1                                                                                          column=e:c1, timestamp=10, value=value
1 row(s) in 0.0650 seconds

KEEP_DELETED_CELLS is to avoid removing Cells from HBase when the only reason to remove them is the delete marker. So with KEEP_DELETED_CELLS enabled deleted cells would get removed if either you write more versions than the configured max, or you have a TTL and Cells are in excess of the configured timeout, etc.

43. Secondary Indexes and Alternate Query Paths

This section could also be titled "what if my table rowkey looks like this but I also want to query my table like that." A common example on the dist-list is where a row-key is of the format "user-timestamp" but there are reporting requirements on activity across users for certain time ranges. Thus, selecting by user is easy because it is in the lead position of the key, but time is not.

There is no single answer on the best way to handle this because it depends on…​

  • Number of users

  • Data size and data arrival rate

  • Flexibility of reporting requirements (e.g., completely ad-hoc date selection vs. pre-configured ranges)

  • Desired execution speed of query (e.g., 90 seconds may be reasonable to some for an ad-hoc report, whereas it may be too long for others)

and solutions are also influenced by the size of the cluster and how much processing power you have to throw at the solution. Common techniques are in sub-sections below. This is a comprehensive, but not exhaustive, list of approaches.

It should not be a surprise that secondary indexes require additional cluster space and processing. This is precisely what happens in an RDBMS because the act of creating an alternate index requires both space and processing cycles to update. RDBMS products are more advanced in this regard to handle alternative index management out of the box. However, HBase scales better at larger data volumes, so this is a feature trade-off.

Pay attention to Apache HBase Performance Tuning when implementing any of these approaches.

Additionally, see the David Butler response in this dist-list thread HBase, mail # user - Stargate+hbase

43.1. Filter Query

Depending on the case, it may be appropriate to use Client Request Filters. In this case, no secondary index is created. However, don’t try a full-scan on a large table like this from an application (i.e., single-threaded client).

43.2. Periodic-Update Secondary Index

A secondary index could be created in another table which is periodically updated via a MapReduce job. The job could be executed intra-day, but depending on load-strategy it could still potentially be out of sync with the main data table.

See mapreduce.example.readwrite for more information.

43.3. Dual-Write Secondary Index

Another strategy is to build the secondary index while publishing data to the cluster (e.g., write to data table, write to index table). If this is approach is taken after a data table already exists, then bootstrapping will be needed for the secondary index with a MapReduce job (see secondary.indexes.periodic).

43.4. Summary Tables

Where time-ranges are very wide (e.g., year-long report) and where the data is voluminous, summary tables are a common approach. These would be generated with MapReduce jobs into another table.

See mapreduce.example.summary for more information.

43.5. Coprocessor Secondary Index

Coprocessors act like RDBMS triggers. These were added in 0.92. For more information, see coprocessors

44. Constraints

HBase currently supports 'constraints' in traditional (SQL) database parlance. The advised usage for Constraints is in enforcing business rules for attributes in the table (e.g. make sure values are in the range 1-10). Constraints could also be used to enforce referential integrity, but this is strongly discouraged as it will dramatically decrease the write throughput of the tables where integrity checking is enabled. Extensive documentation on using Constraints can be found at Constraint since version 0.94.

45. Schema Design Case Studies

The following will describe some typical data ingestion use-cases with HBase, and how the rowkey design and construction can be approached. Note: this is just an illustration of potential approaches, not an exhaustive list. Know your data, and know your processing requirements.

It is highly recommended that you read the rest of the HBase and Schema Design first, before reading these case studies.

The following case studies are described:

  • Log Data / Timeseries Data

  • Log Data / Timeseries on Steroids

  • Customer/Order

  • Tall/Wide/Middle Schema Design

  • List Data

45.1. Case Study - Log Data and Timeseries Data

Assume that the following data elements are being collected.

  • Hostname

  • Timestamp

  • Log event

  • Value/message

We can store them in an HBase table called LOG_DATA, but what will the rowkey be? From these attributes the rowkey will be some combination of hostname, timestamp, and log-event - but what specifically?

45.1.1. Timestamp In The Rowkey Lead Position

The rowkey [timestamp][hostname][log-event] suffers from the monotonically increasing rowkey problem described in Monotonically Increasing Row Keys/Timeseries Data.

There is another pattern frequently mentioned in the dist-lists about "bucketing" timestamps, by performing a mod operation on the timestamp. If time-oriented scans are important, this could be a useful approach. Attention must be paid to the number of buckets, because this will require the same number of scans to return results.

long bucket = timestamp % numBuckets;

to construct:


As stated above, to select data for a particular timerange, a Scan will need to be performed for each bucket. 100 buckets, for example, will provide a wide distribution in the keyspace but it will require 100 Scans to obtain data for a single timestamp, so there are trade-offs.

45.1.2. Host In The Rowkey Lead Position

The rowkey [hostname][log-event][timestamp] is a candidate if there is a large-ish number of hosts to spread the writes and reads across the keyspace. This approach would be useful if scanning by hostname was a priority.

45.1.3. Timestamp, or Reverse Timestamp?

If the most important access path is to pull most recent events, then storing the timestamps as reverse-timestamps (e.g., timestamp = Long.MAX_VALUE – timestamp) will create the property of being able to do a Scan on [hostname][log-event] to obtain the quickly obtain the most recently captured events.

Neither approach is wrong, it just depends on what is most appropriate for the situation.

Reverse Scan API

HBASE-4811 implements an API to scan a table or a range within a table in reverse, reducing the need to optimize your schema for forward or reverse scanning. This feature is available in HBase 0.98 and later. See https://hbase.apache.org/apidocs/org/apache/hadoop/hbase/client/Scan.html#setReversed%28boolean for more information.

45.1.4. Variable Length or Fixed Length Rowkeys?

It is critical to remember that rowkeys are stamped on every column in HBase. If the hostname is a and the event type is e1 then the resulting rowkey would be quite small. However, what if the ingested hostname is myserver1.mycompany.com and the event type is com.package1.subpackage2.subsubpackage3.ImportantService?

It might make sense to use some substitution in the rowkey. There are at least two approaches: hashed and numeric. In the Hostname In The Rowkey Lead Position example, it might look like this:

Composite Rowkey With Hashes:

  • [MD5 hash of hostname] = 16 bytes

  • [MD5 hash of event-type] = 16 bytes

  • [timestamp] = 8 bytes

Composite Rowkey With Numeric Substitution:

For this approach another lookup table would be needed in addition to LOG_DATA, called LOG_TYPES. The rowkey of LOG_TYPES would be:

  • [type] (e.g., byte indicating hostname vs. event-type)

  • [bytes] variable length bytes for raw hostname or event-type.

A column for this rowkey could be a long with an assigned number, which could be obtained by using an HBase counter.

So the resulting composite rowkey would be:

  • [substituted long for hostname] = 8 bytes

  • [substituted long for event type] = 8 bytes

  • [timestamp] = 8 bytes

In either the Hash or Numeric substitution approach, the raw values for hostname and event-type can be stored as columns.

45.2. Case Study - Log Data and Timeseries Data on Steroids

This effectively is the OpenTSDB approach. What OpenTSDB does is re-write data and pack rows into columns for certain time-periods. For a detailed explanation, see: http://opentsdb.net/schema.html, and Lessons Learned from OpenTSDB from HBaseCon2012.

But this is how the general concept works: data is ingested, for example, in this manner…​


with separate rowkeys for each detailed event, but is re-written like this…​


and each of the above events are converted into columns stored with a time-offset relative to the beginning timerange (e.g., every 5 minutes). This is obviously a very advanced processing technique, but HBase makes this possible.

45.3. Case Study - Customer/Order

Assume that HBase is used to store customer and order information. There are two core record-types being ingested: a Customer record type, and Order record type.

The Customer record type would include all the things that you’d typically expect:

  • Customer number

  • Customer name

  • Address (e.g., city, state, zip)

  • Phone numbers, etc.

The Order record type would include things like:

  • Customer number

  • Order number

  • Sales date

  • A series of nested objects for shipping locations and line-items (see Order Object Design for details)

Assuming that the combination of customer number and sales order uniquely identify an order, these two attributes will compose the rowkey, and specifically a composite key such as:

[customer number][order number]

for an ORDER table. However, there are more design decisions to make: are the raw values the best choices for rowkeys?

The same design questions in the Log Data use-case confront us here. What is the keyspace of the customer number, and what is the format (e.g., numeric? alphanumeric?) As it is advantageous to use fixed-length keys in HBase, as well as keys that can support a reasonable spread in the keyspace, similar options appear:

Composite Rowkey With Hashes:

  • [MD5 of customer number] = 16 bytes

  • [MD5 of order number] = 16 bytes

Composite Numeric/Hash Combo Rowkey:

  • [substituted long for customer number] = 8 bytes

  • [MD5 of order number] = 16 bytes

45.3.1. Single Table? Multiple Tables?

A traditional design approach would have separate tables for CUSTOMER and SALES. Another option is to pack multiple record types into a single table (e.g., CUSTOMER++).

Customer Record Type Rowkey:

  • [customer-id]

  • [type] = type indicating `1' for customer record type

Order Record Type Rowkey:

  • [customer-id]

  • [type] = type indicating `2' for order record type

  • [order]

The advantage of this particular CUSTOMER++ approach is that organizes many different record-types by customer-id (e.g., a single scan could get you everything about that customer). The disadvantage is that it’s not as easy to scan for a particular record-type.

45.3.2. Order Object Design

Now we need to address how to model the Order object. Assume that the class structure is as follows:


(an Order can have multiple ShippingLocations


(a ShippingLocation can have multiple LineItems

there are multiple options on storing this data.

Completely Normalized

With this approach, there would be separate tables for ORDER, SHIPPING_LOCATION, and LINE_ITEM.

The ORDER table’s rowkey was described above: schema.casestudies.custorder

The SHIPPING_LOCATION’s composite rowkey would be something like this:

  • [order-rowkey]

  • [shipping location number] (e.g., 1st location, 2nd, etc.)

The LINE_ITEM table’s composite rowkey would be something like this:

  • [order-rowkey]

  • [shipping location number] (e.g., 1st location, 2nd, etc.)

  • [line item number] (e.g., 1st lineitem, 2nd, etc.)

Such a normalized model is likely to be the approach with an RDBMS, but that’s not your only option with HBase. The cons of such an approach is that to retrieve information about any Order, you will need:

  • Get on the ORDER table for the Order

  • Scan on the SHIPPING_LOCATION table for that order to get the ShippingLocation instances

  • Scan on the LINE_ITEM for each ShippingLocation

granted, this is what an RDBMS would do under the covers anyway, but since there are no joins in HBase you’re just more aware of this fact.

Single Table With Record Types

With this approach, there would exist a single table ORDER that would contain

The Order rowkey was described above: schema.casestudies.custorder

  • [order-rowkey]

  • [ORDER record type]

The ShippingLocation composite rowkey would be something like this:

  • [order-rowkey]

  • [SHIPPING record type]

  • [shipping location number] (e.g., 1st location, 2nd, etc.)

The LineItem composite rowkey would be something like this:

  • [order-rowkey]

  • [LINE record type]

  • [shipping location number] (e.g., 1st location, 2nd, etc.)

  • [line item number] (e.g., 1st lineitem, 2nd, etc.)


A variant of the Single Table With Record Types approach is to denormalize and flatten some of the object hierarchy, such as collapsing the ShippingLocation attributes onto each LineItem instance.

The LineItem composite rowkey would be something like this:

  • [order-rowkey]

  • [LINE record type]

  • [line item number] (e.g., 1st lineitem, 2nd, etc., care must be taken that there are unique across the entire order)

and the LineItem columns would be something like this:

  • itemNumber

  • quantity

  • price

  • shipToLine1 (denormalized from ShippingLocation)

  • shipToLine2 (denormalized from ShippingLocation)

  • shipToCity (denormalized from ShippingLocation)

  • shipToState (denormalized from ShippingLocation)

  • shipToZip (denormalized from ShippingLocation)

The pros of this approach include a less complex object hierarchy, but one of the cons is that updating gets more complicated in case any of this information changes.

Object BLOB

With this approach, the entire Order object graph is treated, in one way or another, as a BLOB. For example, the ORDER table’s rowkey was described above: schema.casestudies.custorder, and a single column called "order" would contain an object that could be deserialized that contained a container Order, ShippingLocations, and LineItems.

There are many options here: JSON, XML, Java Serialization, Avro, Hadoop Writables, etc. All of them are variants of the same approach: encode the object graph to a byte-array. Care should be taken with this approach to ensure backward compatibility in case the object model changes such that older persisted structures can still be read back out of HBase.

Pros are being able to manage complex object graphs with minimal I/O (e.g., a single HBase Get per Order in this example), but the cons include the aforementioned warning about backward compatibility of serialization, language dependencies of serialization (e.g., Java Serialization only works with Java clients), the fact that you have to deserialize the entire object to get any piece of information inside the BLOB, and the difficulty in getting frameworks like Hive to work with custom objects like this.

45.4. Case Study - "Tall/Wide/Middle" Schema Design Smackdown

This section will describe additional schema design questions that appear on the dist-list, specifically about tall and wide tables. These are general guidelines and not laws - each application must consider its own needs.

45.4.1. Rows vs. Versions

A common question is whether one should prefer rows or HBase’s built-in-versioning. The context is typically where there are "a lot" of versions of a row to be retained (e.g., where it is significantly above the HBase default of 1 max versions). The rows-approach would require storing a timestamp in some portion of the rowkey so that they would not overwrite with each successive update.

Preference: Rows (generally speaking).

45.4.2. Rows vs. Columns

Another common question is whether one should prefer rows or columns. The context is typically in extreme cases of wide tables, such as having 1 row with 1 million attributes, or 1 million rows with 1 columns apiece.

Preference: Rows (generally speaking). To be clear, this guideline is in the context is in extremely wide cases, not in the standard use-case where one needs to store a few dozen or hundred columns. But there is also a middle path between these two options, and that is "Rows as Columns."

45.4.3. Rows as Columns

The middle path between Rows vs. Columns is packing data that would be a separate row into columns, for certain rows. OpenTSDB is the best example of this case where a single row represents a defined time-range, and then discrete events are treated as columns. This approach is often more complex, and may require the additional complexity of re-writing your data, but has the advantage of being I/O efficient. For an overview of this approach, see schema.casestudies.log-steroids.

45.5. Case Study - List Data

The following is an exchange from the user dist-list regarding a fairly common question: how to handle per-user list data in Apache HBase.


We’re looking at how to store a large amount of (per-user) list data in HBase, and we were trying to figure out what kind of access pattern made the most sense. One option is store the majority of the data in a key, so we could have something like:

<FixedWidthUserName><FixedWidthValueId1>:"" (no value)
<FixedWidthUserName><FixedWidthValueId2>:"" (no value)
<FixedWidthUserName><FixedWidthValueId3>:"" (no value)

The other option we had was to do this entirely using:


where each row would contain multiple values. So in one case reading the first thirty values would be:

scan { STARTROW => 'FixedWidthUsername' LIMIT => 30}

And in the second case it would be

get 'FixedWidthUserName\x00\x00\x00\x00'

The general usage pattern would be to read only the first 30 values of these lists, with infrequent access reading deeper into the lists. Some users would have ⇐ 30 total values in these lists, and some users would have millions (i.e. power-law distribution)

The single-value format seems like it would take up more space on HBase, but would offer some improved retrieval / pagination flexibility. Would there be any significant performance advantages to be able to paginate via gets vs paginating with scans?

My initial understanding was that doing a scan should be faster if our paging size is unknown (and caching is set appropriately), but that gets should be faster if we’ll always need the same page size. I’ve ended up hearing different people tell me opposite things about performance. I assume the page sizes would be relatively consistent, so for most use cases we could guarantee that we only wanted one page of data in the fixed-page-length case. I would also assume that we would have infrequent updates, but may have inserts into the middle of these lists (meaning we’d need to update all subsequent rows).

Thanks for help / suggestions / follow-up questions.

  • ANSWER *

If I understand you correctly, you’re ultimately trying to store triples in the form "user, valueid, value", right? E.g., something like:

"user123, firstname, Paul",
"user234, lastname, Smith"

(But the usernames are fixed width, and the valueids are fixed width).

And, your access pattern is along the lines of: "for user X, list the next 30 values, starting with valueid Y". Is that right? And these values should be returned sorted by valueid?

The tl;dr version is that you should probably go with one row per user+value, and not build a complicated intra-row pagination scheme on your own unless you’re really sure it is needed.

Your two options mirror a common question people have when designing HBase schemas: should I go "tall" or "wide"? Your first schema is "tall": each row represents one value for one user, and so there are many rows in the table for each user; the row key is user + valueid, and there would be (presumably) a single column qualifier that means "the value". This is great if you want to scan over rows in sorted order by row key (thus my question above, about whether these ids are sorted correctly). You can start a scan at any user+valueid, read the next 30, and be done. What you’re giving up is the ability to have transactional guarantees around all the rows for one user, but it doesn’t sound like you need that. Doing it this way is generally recommended (see here http://hbase.apache.org/book.html#schema.smackdown).

Your second option is "wide": you store a bunch of values in one row, using different qualifiers (where the qualifier is the valueid). The simple way to do that would be to just store ALL values for one user in a single row. I’m guessing you jumped to the "paginated" version because you’re assuming that storing millions of columns in a single row would be bad for performance, which may or may not be true; as long as you’re not trying to do too much in a single request, or do things like scanning over and returning all of the cells in the row, it shouldn’t be fundamentally worse. The client has methods that allow you to get specific slices of columns.

Note that neither case fundamentally uses more disk space than the other; you’re just "shifting" part of the identifying information for a value either to the left (into the row key, in option one) or to the right (into the column qualifiers in option 2). Under the covers, every key/value still stores the whole row key, and column family name. (If this is a bit confusing, take an hour and watch Lars George’s excellent video about understanding HBase schema design: http://www.youtube.com/watch?v=_HLoH_PgrLk).

A manually paginated version has lots more complexities, as you note, like having to keep track of how many things are in each page, re-shuffling if new values are inserted, etc. That seems significantly more complex. It might have some slight speed advantages (or disadvantages!) at extremely high throughput, and the only way to really know that would be to try it out. If you don’t have time to build it both ways and compare, my advice would be to start with the simplest option (one row per user+value). Start simple and iterate! :)

46. Operational and Performance Configuration Options

See the Performance section perf.schema for more information operational and performance schema design options, such as Bloom Filters, Table-configured regionsizes, compression, and blocksizes.

HBase and MapReduce

Apache MapReduce is a software framework used to analyze large amounts of data, and is the framework used most often with Apache Hadoop. MapReduce itself is out of the scope of this document. A good place to get started with MapReduce is http://hadoop.apache.org/docs/r2.6.0/hadoop-mapreduce-client/hadoop-mapreduce-client-core/MapReduceTutorial.html. MapReduce version 2 (MR2)is now part of YARN.

This chapter discusses specific configuration steps you need to take to use MapReduce on data within HBase. In addition, it discusses other interactions and issues between HBase and MapReduce jobs. Finally, it discusses Cascading, an alternative API for MapReduce.

mapred and mapreduce

There are two mapreduce packages in HBase as in MapReduce itself: org.apache.hadoop.hbase.mapred and org.apache.hadoop.hbase.mapreduce. The former does old-style API and the latter the new style. The latter has more facility though you can usually find an equivalent in the older package. Pick the package that goes with your MapReduce deploy. When in doubt or starting over, pick the org.apache.hadoop.hbase.mapreduce. In the notes below, we refer to o.a.h.h.mapreduce but replace with the o.a.h.h.mapred if that is what you are using.

47. HBase, MapReduce, and the CLASSPATH

By default, MapReduce jobs deployed to a MapReduce cluster do not have access to either the HBase configuration under $HBASE_CONF_DIR or the HBase classes.

To give the MapReduce jobs the access they need, you could add hbase-site.xml to $HADOOP_HOME/conf and add HBase jars to the $HADOOP_HOME/lib directory. You would then need to copy these changes across your cluster. Or you can edit $HADOOP_HOME/conf/hadoop-env.sh and add them to the HADOOP_CLASSPATH variable. However, this approach is not recommended because it will pollute your Hadoop install with HBase references. It also requires you to restart the Hadoop cluster before Hadoop can use the HBase data.

The recommended approach is to let HBase add its dependency jars itself and use HADOOP_CLASSPATH or -libjars.

Since HBase 0.90.x, HBase adds its dependency JARs to the job configuration itself. The dependencies only need to be available on the local CLASSPATH. The following example runs the bundled HBase RowCounter MapReduce job against a table named usertable. If you have not set the environment variables expected in the command (the parts prefixed by a $ sign and surrounded by curly braces), you can use the actual system paths instead. Be sure to use the correct version of the HBase JAR for your system. The backticks (` symbols) cause the shell to execute the sub-commands, setting the output of hbase classpath (the command to dump HBase CLASSPATH) to HADOOP_CLASSPATH. This example assumes you use a BASH-compatible shell.

$ HADOOP_CLASSPATH=`${HBASE_HOME}/bin/hbase classpath` ${HADOOP_HOME}/bin/hadoop jar ${HBASE_HOME}/lib/hbase-server-VERSION.jar rowcounter usertable

When the command runs, internally, the HBase JAR finds the dependencies it needs for ZooKeeper, Guava, and its other dependencies on the passed HADOOP_CLASSPATH and adds the JARs to the MapReduce job configuration. See the source at TableMapReduceUtil#addDependencyJars(org.apache.hadoop.mapreduce.Job) for how this is done.

The command hbase mapredcp can also help you dump the CLASSPATH entries required by MapReduce, which are the same jars TableMapReduceUtil#addDependencyJars would add. You can add them together with HBase conf directory to HADOOP_CLASSPATH. For jobs that do not package their dependencies or call TableMapReduceUtil#addDependencyJars, the following command structure is necessary:

$ HADOOP_CLASSPATH=`${HBASE_HOME}/bin/hbase mapredcp`:${HBASE_HOME}/conf hadoop jar MyApp.jar MyJobMainClass -libjars $(${HBASE_HOME}/bin/hbase mapredcp | tr ':' ',') ...

The example may not work if you are running HBase from its build directory rather than an installed location. You may see an error like the following:

java.lang.RuntimeException: java.lang.ClassNotFoundException: org.apache.hadoop.hbase.mapreduce.RowCounter$RowCounterMapper

If this occurs, try modifying the command as follows, so that it uses the HBase JARs from the target/ directory within the build environment.

$ HADOOP_CLASSPATH=${HBASE_BUILD_HOME}/hbase-server/target/hbase-server-VERSION-SNAPSHOT.jar:`${HBASE_BUILD_HOME}/bin/hbase classpath` ${HADOOP_HOME}/bin/hadoop jar ${HBASE_BUILD_HOME}/hbase-server/target/hbase-server-VERSION-SNAPSHOT.jar rowcounter usertable
Notice to MapReduce users of HBase between 0.96.1 and 0.98.4

Some MapReduce jobs that use HBase fail to launch. The symptom is an exception similar to the following:

Exception in thread "main" java.lang.IllegalAccessError: class
    com.google.protobuf.ZeroCopyLiteralByteString cannot access its superclass
    at java.lang.ClassLoader.defineClass1(Native Method)
    at java.lang.ClassLoader.defineClass(ClassLoader.java:792)
    at java.security.SecureClassLoader.defineClass(SecureClassLoader.java:142)
    at java.net.URLClassLoader.defineClass(URLClassLoader.java:449)
    at java.net.URLClassLoader.access$100(URLClassLoader.java:71)
    at java.net.URLClassLoader$1.run(URLClassLoader.java:361)
    at java.net.URLClassLoader$1.run(URLClassLoader.java:355)
    at java.security.AccessController.doPrivileged(Native Method)
    at java.net.URLClassLoader.findClass(URLClassLoader.java:354)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:357)

This is caused by an optimization introduced in HBASE-9867 that inadvertently introduced a classloader dependency.

This affects both jobs using the -libjars option and "fat jar," those which package their runtime dependencies in a nested lib folder.

In order to satisfy the new classloader requirements, hbase-protocol.jar must be included in Hadoop’s classpath. See HBase, MapReduce, and the CLASSPATH for current recommendations for resolving classpath errors. The following is included for historical purposes.

This can be resolved system-wide by including a reference to the hbase-protocol.jar in Hadoop’s lib directory, via a symlink or by copying the jar into the new location.

This can also be achieved on a per-job launch basis by including it in the HADOOP_CLASSPATH environment variable at job submission time. When launching jobs that package their dependencies, all three of the following job launching commands satisfy this requirement:

$ HADOOP_CLASSPATH=/path/to/hbase-protocol.jar:/path/to/hbase/conf hadoop jar MyJob.jar MyJobMainClass
$ HADOOP_CLASSPATH=$(hbase mapredcp):/path/to/hbase/conf hadoop jar MyJob.jar MyJobMainClass
$ HADOOP_CLASSPATH=$(hbase classpath) hadoop jar MyJob.jar MyJobMainClass

For jars that do not package their dependencies, the following command structure is necessary:

$ HADOOP_CLASSPATH=$(hbase mapredcp):/etc/hbase/conf hadoop jar MyApp.jar MyJobMainClass -libjars $(hbase mapredcp | tr ':' ',') ...

See also HBASE-10304 for further discussion of this issue.

48. MapReduce Scan Caching

TableMapReduceUtil now restores the option to set scanner caching (the number of rows which are cached before returning the result to the client) on the Scan object that is passed in. This functionality was lost due to a bug in HBase 0.95 (HBASE-11558), which is fixed for HBase 0.98.5 and 0.96.3. The priority order for choosing the scanner caching is as follows:

  1. Caching settings which are set on the scan object.

  2. Caching settings which are specified via the configuration option hbase.client.scanner.caching, which can either be set manually in hbase-site.xml or via the helper method TableMapReduceUtil.setScannerCaching().

  3. The default value HConstants.DEFAULT_HBASE_CLIENT_SCANNER_CACHING, which is set to 100.

Optimizing the caching settings is a balance between the time the client waits for a result and the number of sets of results the client needs to receive. If the caching setting is too large, the client could end up waiting for a long time or the request could even time out. If the setting is too small, the scan needs to return results in several pieces. If you think of the scan as a shovel, a bigger cache setting is analogous to a bigger shovel, and a smaller cache setting is equivalent to more shoveling in order to fill the bucket.

The list of priorities mentioned above allows you to set a reasonable default, and override it for specific operations.

See the API documentation for Scan for more details.

49. Bundled HBase MapReduce Jobs

The HBase JAR also serves as a Driver for some bundled MapReduce jobs. To learn about the bundled MapReduce jobs, run the following command.

$ ${HADOOP_HOME}/bin/hadoop jar ${HBASE_HOME}/hbase-server-VERSION.jar
An example program must be given as the first argument.
Valid program names are:
  copytable: Export a table from local cluster to peer cluster
  completebulkload: Complete a bulk data load.
  export: Write table data to HDFS.
  import: Import data written by Export.
  importtsv: Import data in TSV format.
  rowcounter: Count rows in HBase table

Each of the valid program names are bundled MapReduce jobs. To run one of the jobs, model your command after the following example.

$ ${HADOOP_HOME}/bin/hadoop jar ${HBASE_HOME}/hbase-server-VERSION.jar rowcounter myTable

50. HBase as a MapReduce Job Data Source and Data Sink

HBase can be used as a data source, TableInputFormat, and data sink, TableOutputFormat or MultiTableOutputFormat, for MapReduce jobs. Writing MapReduce jobs that read or write HBase, it is advisable to subclass TableMapper and/or TableReducer. See the do-nothing pass-through classes IdentityTableMapper and IdentityTableReducer for basic usage. For a more involved example, see RowCounter or review the org.apache.hadoop.hbase.mapreduce.TestTableMapReduce unit test.

If you run MapReduce jobs that use HBase as source or sink, need to specify source and sink table and column names in your configuration.

When you read from HBase, the TableInputFormat requests the list of regions from HBase and makes a map, which is either a map-per-region or mapreduce.job.maps map, whichever is smaller. If your job only has two maps, raise mapreduce.job.maps to a number greater than the number of regions. Maps will run on the adjacent TaskTracker/NodeManager if you are running a TaskTracer/NodeManager and RegionServer per node. When writing to HBase, it may make sense to avoid the Reduce step and write back into HBase from within your map. This approach works when your job does not need the sort and collation that MapReduce does on the map-emitted data. On insert, HBase 'sorts' so there is no point double-sorting (and shuffling data around your MapReduce cluster) unless you need to. If you do not need the Reduce, your map might emit counts of records processed for reporting at the end of the job, or set the number of Reduces to zero and use TableOutputFormat. If running the Reduce step makes sense in your case, you should typically use multiple reducers so that load is spread across the HBase cluster.

A new HBase partitioner, the HRegionPartitioner, can run as many reducers the number of existing regions. The HRegionPartitioner is suitable when your table is large and your upload will not greatly alter the number of existing regions upon completion. Otherwise use the default partitioner.

51. Writing HFiles Directly During Bulk Import

If you are importing into a new table, you can bypass the HBase API and write your content directly to the filesystem, formatted into HBase data files (HFiles). Your import will run faster, perhaps an order of magnitude faster. For more on how this mechanism works, see Bulk Loading.

52. RowCounter Example

The included RowCounter MapReduce job uses TableInputFormat and does a count of all rows in the specified table. To run it, use the following command:

$ ./bin/hadoop jar hbase-X.X.X.jar

This will invoke the HBase MapReduce Driver class. Select rowcounter from the choice of jobs offered. This will print rowcounter usage advice to standard output. Specify the tablename, column to count, and output directory. If you have classpath errors, see HBase, MapReduce, and the CLASSPATH.

53. Map-Task Splitting

53.1. The Default HBase MapReduce Splitter

When TableInputFormat is used to source an HBase table in a MapReduce job, its splitter will make a map task for each region of the table. Thus, if there are 100 regions in the table, there will be 100 map-tasks for the job - regardless of how many column families are selected in the Scan.

53.2. Custom Splitters

For those interested in implementing custom splitters, see the method getSplits in TableInputFormatBase. That is where the logic for map-task assignment resides.

54. HBase MapReduce Examples

54.1. HBase MapReduce Read Example

The following is an example of using HBase as a MapReduce source in read-only manner. Specifically, there is a Mapper instance but no Reducer, and nothing is being emitted from the Mapper. The job would be defined as follows…​

Configuration config = HBaseConfiguration.create();
Job job = new Job(config, "ExampleRead");
job.setJarByClass(MyReadJob.class);     // class that contains mapper

Scan scan = new Scan();
scan.setCaching(500);        // 1 is the default in Scan, which will be bad for MapReduce jobs
scan.setCacheBlocks(false);  // don't set to true for MR jobs
// set other scan attrs

  tableName,        // input HBase table name
  scan,             // Scan instance to control CF and attribute selection
  MyMapper.class,   // mapper
  null,             // mapper output key
  null,             // mapper output value
job.setOutputFormatClass(NullOutputFormat.class);   // because we aren't emitting anything from mapper

boolean b = job.waitForCompletion(true);
if (!b) {
  throw new IOException("error with job!");

…​and the mapper instance would extend TableMapper…​

public static class MyMapper extends TableMapper<Text, Text> {

  public void map(ImmutableBytesWritable row, Result value, Context context) throws InterruptedException, IOException {
    // process data for the row from the Result instance.

54.2. HBase MapReduce Read/Write Example

The following is an example of using HBase both as a source and as a sink with MapReduce. This example will simply copy data from one table to another.

Configuration config = HBaseConfiguration.create();
Job job = new Job(config,"ExampleReadWrite");
job.setJarByClass(MyReadWriteJob.class);    // class that contains mapper

Scan scan = new Scan();
scan.setCaching(500);        // 1 is the default in Scan, which will be bad for MapReduce jobs
scan.setCacheBlocks(false);  // don't set to true for MR jobs
// set other scan attrs

  sourceTable,      // input table
  scan,             // Scan instance to control CF and attribute selection
  MyMapper.class,   // mapper class
  null,             // mapper output key
  null,             // mapper output value
  targetTable,      // output table
  null,             // reducer class

boolean b = job.waitForCompletion(true);
if (!b) {
    throw new IOException("error with job!");

An explanation is required of what TableMapReduceUtil is doing, especially with the reducer. TableOutputFormat is being used as the outputFormat class, and several parameters are being set on the config (e.g., TableOutputFormat.OUTPUT_TABLE), as well as setting the reducer output key to ImmutableBytesWritable and reducer value to Writable. These could be set by the programmer on the job and conf, but TableMapReduceUtil tries to make things easier.

The following is the example mapper, which will create a Put and matching the input Result and emit it. Note: this is what the CopyTable utility does.

public static class MyMapper extends TableMapper<ImmutableBytesWritable, Put>  {

  public void map(ImmutableBytesWritable row, Result value, Context context) throws IOException, InterruptedException {
    // this example is just copying the data from the source table...
      context.write(row, resultToPut(row,value));

    private static Put resultToPut(ImmutableBytesWritable key, Result result) throws IOException {
      Put put = new Put(key.get());
      for (KeyValue kv : result.raw()) {
      return put;

There isn’t actually a reducer step, so TableOutputFormat takes care of sending the Put to the target table.

This is just an example, developers could choose not to use TableOutputFormat and connect to the target table themselves.

54.3. HBase MapReduce Read/Write Example With Multi-Table Output

TODO: example for MultiTableOutputFormat.

54.4. HBase MapReduce Summary to HBase Example

The following example uses HBase as a MapReduce source and sink with a summarization step. This example will count the number of distinct instances of a value in a table and write those summarized counts in another table.

Configuration config = HBaseConfiguration.create();
Job job = new Job(config,"ExampleSummary");
job.setJarByClass(MySummaryJob.class);     // class that contains mapper and reducer

Scan scan = new Scan();
scan.setCaching(500);        // 1 is the default in Scan, which will be bad for MapReduce jobs
scan.setCacheBlocks(false);  // don't set to true for MR jobs
// set other scan attrs

  sourceTable,        // input table
  scan,               // Scan instance to control CF and attribute selection
  MyMapper.class,     // mapper class
  Text.class,         // mapper output key
  IntWritable.class,  // mapper output value
  targetTable,        // output table
  MyTableReducer.class,    // reducer class
job.setNumReduceTasks(1);   // at least one, adjust as required

boolean b = job.waitForCompletion(true);
if (!b) {
  throw new IOException("error with job!");

In this example mapper a column with a String-value is chosen as the value to summarize upon. This value is used as the key to emit from the mapper, and an IntWritable represents an instance counter.

public static class MyMapper extends TableMapper<Text, IntWritable>  {
  public static final byte[] CF = "cf".getBytes();
  public static final byte[] ATTR1 = "attr1".getBytes();

  private final IntWritable ONE = new IntWritable(1);
  private Text text = new Text();

  public void map(ImmutableBytesWritable row, Result value, Context context) throws IOException, InterruptedException {
    String val = new String(value.getValue(CF, ATTR1));
    text.set(val);     // we can only emit Writables...
    context.write(text, ONE);

In the reducer, the "ones" are counted (just like any other MR example that does this), and then emits a Put.

public static class MyTableReducer extends TableReducer<Text, IntWritable, ImmutableBytesWritable>  {
  public static final byte[] CF = "cf".getBytes();
  public static final byte[] COUNT = "count".getBytes();

  public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
    int i = 0;
    for (IntWritable val : values) {
      i += val.get();
    Put put = new Put(Bytes.toBytes(key.toString()));
    put.add(CF, COUNT, Bytes.toBytes(i));

    context.write(null, put);

54.5. HBase MapReduce Summary to File Example

This very similar to the summary example above, with exception that this is using HBase as a MapReduce source but HDFS as the sink. The differences are in the job setup and in the reducer. The mapper remains the same.

Configuration config = HBaseConfiguration.create();
Job job = new Job(config,"ExampleSummaryToFile");
job.setJarByClass(MySummaryFileJob.class);     // class that contains mapper and reducer

Scan scan = new Scan();
scan.setCaching(500);        // 1 is the default in Scan, which will be bad for MapReduce jobs
scan.setCacheBlocks(false);  // don't set to true for MR jobs
// set other scan attrs

  sourceTable,        // input table
  scan,               // Scan instance to control CF and attribute selection
  MyMapper.class,     // mapper class
  Text.class,         // mapper output key
  IntWritable.class,  // mapper output value
job.setReducerClass(MyReducer.class);    // reducer class
job.setNumReduceTasks(1);    // at least one, adjust as required
FileOutputFormat.setOutputPath(job, new Path("/tmp/mr/mySummaryFile"));  // adjust directories as required

boolean b = job.waitForCompletion(true);
if (!b) {
  throw new IOException("error with job!");

As stated above, the previous Mapper can run unchanged with this example. As for the Reducer, it is a "generic" Reducer instead of extending TableMapper and emitting Puts.

public static class MyReducer extends Reducer<Text, IntWritable, Text, IntWritable>  {

  public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
    int i = 0;
    for (IntWritable val : values) {
      i += val.get();
    context.write(key, new IntWritable(i));

54.6. HBase MapReduce Summary to HBase Without Reducer

It is also possible to perform summaries without a reducer - if you use HBase as the reducer.

An HBase target table would need to exist for the job summary. The Table method incrementColumnValue would be used to atomically increment values. From a performance perspective, it might make sense to keep a Map of values with their values to be incremented for each map-task, and make one update per key at during the cleanup method of the mapper. However, your mileage may vary depending on the number of rows to be processed and unique keys.

In the end, the summary results are in HBase.

54.7. HBase MapReduce Summary to RDBMS

Sometimes it is more appropriate to generate summaries to an RDBMS. For these cases, it is possible to generate summaries directly to an RDBMS via a custom reducer. The setup method can connect to an RDBMS (the connection information can be passed via custom parameters in the context) and the cleanup method can close the connection.

It is critical to understand that number of reducers for the job affects the summarization implementation, and you’ll have to design this into your reducer. Specifically, whether it is designed to run as a singleton (one reducer) or multiple reducers. Neither is right or wrong, it depends on your use-case. Recognize that the more reducers that are assigned to the job, the more simultaneous connections to the RDBMS will be created - this will scale, but only to a point.

public static class MyRdbmsReducer extends Reducer<Text, IntWritable, Text, IntWritable>  {

  private Connection c = null;

  public void setup(Context context) {
    // create DB connection...

  public void reduce(Text key, Iterable<IntWritable> values, Context context) throws IOException, InterruptedException {
    // do summarization
    // in this example the keys are Text, but this is just an example

  public void cleanup(Context context) {
    // close db connection


In the end, the summary results are written to your RDBMS table/s.

55. Accessing Other HBase Tables in a MapReduce Job

Although the framework currently allows one HBase table as input to a MapReduce job, other HBase tables can be accessed as lookup tables, etc., in a MapReduce job via creating an Table instance in the setup method of the Mapper.

public class MyMapper extends TableMapper<Text, LongWritable> {
  private Table myOtherTable;

  public void setup(Context context) {
    // In here create a Connection to the cluster and save it or use the Connection
    // from the existing table
    myOtherTable = connection.getTable("myOtherTable");

  public void map(ImmutableBytesWritable row, Result value, Context context) throws IOException, InterruptedException {
    // process Result...
    // use 'myOtherTable' for lookups

56. Speculative Execution

It is generally advisable to turn off speculative execution for MapReduce jobs that use HBase as a source. This can either be done on a per-Job basis through properties, or on the entire cluster. Especially for longer running jobs, speculative execution will create duplicate map-tasks which will double-write your data to HBase; this is probably not what you want.

See spec.ex for more information.

57. Cascading

Cascading is an alternative API for MapReduce, which actually uses MapReduce, but allows you to write your MapReduce code in a simplified way.

The following example shows a Cascading Flow which "sinks" data into an HBase cluster. The same hBaseTap API could be used to "source" data as well.

// read data from the default filesystem
// emits two fields: "offset" and "line"
Tap source = new Hfs( new TextLine(), inputFileLhs );

// store data in an HBase cluster
// accepts fields "num", "lower", and "upper"
// will automatically scope incoming fields to their proper familyname, "left" or "right"
Fields keyFields = new Fields( "num" );
String[] familyNames = {"left", "right"};
Fields[] valueFields = new Fields[] {new Fields( "lower" ), new Fields( "upper" ) };
Tap hBaseTap = new HBaseTap( "multitable", new HBaseScheme( keyFields, familyNames, valueFields ), SinkMode.REPLACE );

// a simple pipe assembly to parse the input into fields
// a real app would likely chain multiple Pipes together for more complex processing
Pipe parsePipe = new Each( "insert", new Fields( "line" ), new RegexSplitter( new Fields( "num", "lower", "upper" ), " " ) );

// "plan" a cluster executable Flow
// this connects the source Tap and hBaseTap (the sink Tap) to the parsePipe
Flow parseFlow = new FlowConnector( properties ).connect( source, hBaseTap, parsePipe );

// start the flow, and block until complete

// open an iterator on the HBase table we stuffed data into
TupleEntryIterator iterator = parseFlow.openSink();

  // print out each tuple from HBase
  System.out.println( "iterator.next() = " + iterator.next() );


Securing Apache HBase

Reporting Security Bugs
To protect existing HBase installations from exploitation, please do not use JIRA to report security-related bugs. Instead, send your report to the mailing list private@hbase.apache.org, which allows anyone to send messages, but restricts who can read them. Someone on that list will contact you to follow up on your report.

HBase adheres to the Apache Software Foundation’s policy on reported vulnerabilities, available at http://apache.org/security/.

If you wish to send an encrypted report, you can use the GPG details provided for the general ASF security list. This will likely increase the response time to your report.

HBase provides mechanisms to secure various components and aspects of HBase and how it relates to the rest of the Hadoop infrastructure, as well as clients and resources outside Hadoop.

58. Using Secure HTTP (HTTPS) for the Web UI

A default HBase install uses insecure HTTP connections for Web UIs for the master and region servers. To enable secure HTTP (HTTPS) connections instead, set hbase.ssl.enabled to true in hbase-site.xml. This does not change the port used by the Web UI. To change the port for the web UI for a given HBase component, configure that port’s setting in hbase-site.xml. These settings are:

  • hbase.master.info.port

  • hbase.regionserver.info.port

If you enable HTTPS, clients should avoid using the non-secure HTTP connection.

If you enable secure HTTP, clients should connect to HBase using the https:// URL. Clients using the http:// URL will receive an HTTP response of 200, but will not receive any data. The following exception is logged:

javax.net.ssl.SSLException: Unrecognized SSL message, plaintext connection?

This is because the same port is used for HTTP and HTTPS.

HBase uses Jetty for the Web UI. Without modifying Jetty itself, it does not seem possible to configure Jetty to redirect one port to another on the same host. See Nick Dimiduk’s contribution on this Stack Overflow thread for more information. If you know how to fix this without opening a second port for HTTPS, patches are appreciated.

59. Secure Client Access to Apache HBase

Newer releases of Apache HBase (>= 0.92) support optional SASL authentication of clients. See also Matteo Bertozzi’s article on Understanding User Authentication and Authorization in Apache HBase.

This describes how to set up Apache HBase and clients for connection to secure HBase resources.

59.1. Prerequisites

Hadoop Authentication Configuration

To run HBase RPC with strong authentication, you must set hbase.security.authentication to kerberos. In this case, you must also set hadoop.security.authentication to kerberos in core-site.xml. Otherwise, you would be using strong authentication for HBase but not for the underlying HDFS, which would cancel out any benefit.

Kerberos KDC

You need to have a working Kerberos KDC.

59.2. Server-side Configuration for Secure Operation

First, refer to security.prerequisites and ensure that your underlying HDFS configuration is secure.

Add the following to the hbase-site.xml file on every server machine in the cluster:


A full shutdown and restart of HBase service is required when deploying these configuration changes.

59.3. Client-side Configuration for Secure Operation

First, refer to Prerequisites and ensure that your underlying HDFS configuration is secure.

Add the following to the hbase-site.xml file on every client:


The client environment must be logged in to Kerberos from KDC or keytab via the kinit command before communication with the HBase cluster will be possible.

Be advised that if the hbase.security.authentication in the client- and server-side site files do not match, the client will not be able to communicate with the cluster.

Once HBase is configured for secure RPC it is possible to optionally configure encrypted communication. To do so, add the following to the hbase-site.xml file on every client:


This configuration property can also be set on a per-connection basis. Set it in the Configuration supplied to Table:

Configuration conf = HBaseConfiguration.create();
Connection connection = ConnectionFactory.createConnection(conf);
conf.set("hbase.rpc.protection", "privacy");
try (Connection connection = ConnectionFactory.createConnection(conf)) {
  try (Table table = connection.getTable(TableName.valueOf(tablename)) {
  .... do your stuff

Expect a ~10% performance penalty for encrypted communication.

59.4. Client-side Configuration for Secure Operation - Thrift Gateway

Add the following to the hbase-site.xml file for every Thrift gateway:

  <!-- TODO: This may need to be  HTTP/_HOST@<REALM> and _HOST may not work.
   You may have  to put the concrete full hostname.
<!-- Add these if you need to configure a different DNS interface from the default -->

Substitute the appropriate credential and keytab for $USER and $KEYTAB respectively.

In order to use the Thrift API principal to interact with HBase, it is also necessary to add the hbase.thrift.kerberos.principal to the acl table. For example, to give the Thrift API principal, thrift_server, administrative access, a command such as this one will suffice:

grant 'thrift_server', 'RWCA'

For more information about ACLs, please see the Access Control Labels (ACLs) section

The Thrift gateway will authenticate with HBase using the supplied credential. No authentication will be performed by the Thrift gateway itself. All client access via the Thrift gateway will use the Thrift gateway’s credential and have its privilege.

59.5. Configure the Thrift Gateway to Authenticate on Behalf of the Client

Client-side Configuration for Secure Operation - Thrift Gateway describes how to authenticate a Thrift client to HBase using a fixed user. As an alternative, you can configure the Thrift gateway to authenticate to HBase on the client’s behalf, and to access HBase using a proxy user. This was implemented in HBASE-11349 for Thrift 1, and HBASE-11474 for Thrift 2.

Limitations with Thrift Framed Transport

If you use framed transport, you cannot yet take advantage of this feature, because SASL does not work with Thrift framed transport at this time.

To enable it, do the following.

  1. Be sure Thrift is running in secure mode, by following the procedure described in Client-side Configuration for Secure Operation - Thrift Gateway.

  2. Be sure that HBase is configured to allow proxy users, as described in REST Gateway Impersonation Configuration.

  3. In hbase-site.xml for each cluster node running a Thrift gateway, set the property hbase.thrift.security.qop to one of the following three values:

    • privacy - authentication, integrity, and confidentiality checking.

    • integrity - authentication and integrity checking

    • authentication - authentication checking only

  4. Restart the Thrift gateway processes for the changes to take effect. If a node is running Thrift, the output of the jps command will list a ThriftServer process. To stop Thrift on a node, run the command bin/hbase-daemon.sh stop thrift. To start Thrift on a node, run the command bin/hbase-daemon.sh start thrift.

59.6. Configure the Thrift Gateway to Use the doAs Feature

Configure the Thrift Gateway to Authenticate on Behalf of the Client describes how to configure the Thrift gateway to authenticate to HBase on the client’s behalf, and to access HBase using a proxy user. The limitation of this approach is that after the client is initialized with a particular set of credentials, it cannot change these credentials during the session. The doAs feature provides a flexible way to impersonate multiple principals using the same client. This feature was implemented in HBASE-12640 for Thrift 1, but is currently not available for Thrift 2.

To enable the doAs feature, add the following to the hbase-site.xml file for every Thrift gateway:


To allow proxy users when using doAs impersonation, add the following to the hbase-site.xml file for every HBase node:


Take a look at the demo client to get an overall idea of how to use this feature in your client.

59.7. Client-side Configuration for Secure Operation - REST Gateway

Add the following to the hbase-site.xml file for every REST gateway:


Substitute the appropriate credential and keytab for $USER and $KEYTAB respectively.

The REST gateway will authenticate with HBase using the supplied credential.

In order to use the REST API principal to interact with HBase, it is also necessary to add the hbase.rest.kerberos.principal to the acl table. For example, to give the REST API principal, rest_server, administrative access, a command such as this one will suffice:

grant 'rest_server', 'RWCA'

For more information about ACLs, please see the Access Control Labels (ACLs) section

HBase REST gateway supports SPNEGO HTTP authentication for client access to the gateway. To enable REST gateway Kerberos authentication for client access, add the following to the hbase-site.xml file for every REST gateway.

<!-- Add these if you need to configure a different DNS interface from the default -->

Substitute the keytab for HTTP for $KEYTAB.

HBase REST gateway supports different 'hbase.rest.authentication.type': simple, kerberos. You can also implement a custom authentication by implementing Hadoop AuthenticationHandler, then specify the full class name as 'hbase.rest.authentication.type' value. For more information, refer to SPNEGO HTTP authentication.

59.8. REST Gateway Impersonation Configuration

By default, the REST gateway doesn’t support impersonation. It accesses the HBase on behalf of clients as the user configured as in the previous section. To the HBase server, all requests are from the REST gateway user. The actual users are unknown. You can turn on the impersonation support. With impersonation, the REST gateway user is a proxy user. The HBase server knows the actual/real user of each request. So it can apply proper authorizations.

To turn on REST gateway impersonation, we need to configure HBase servers (masters and region servers) to allow proxy users; configure REST gateway to enable impersonation.

To allow proxy users, add the following to the hbase-site.xml file for every HBase server:


Substitute the REST gateway proxy user for $USER, and the allowed group list for $GROUPS.

To enable REST gateway impersonation, add the following to the hbase-site.xml file for every REST gateway.


Substitute the keytab for HTTP for $KEYTAB.

60. Simple User Access to Apache HBase

Newer releases of Apache HBase (>= 0.92) support optional SASL authentication of clients. See also Matteo Bertozzi’s article on Understanding User Authentication and Authorization in Apache HBase.

This describes how to set up Apache HBase and clients for simple user access to HBase resources.

60.1. Simple versus Secure Access

The following section shows how to set up simple user access. Simple user access is not a secure method of operating HBase. This method is used to prevent users from making mistakes. It can be used to mimic the Access Control using on a development system without having to set up Kerberos.

This method is not used to prevent malicious or hacking attempts. To make HBase secure against these types of attacks, you must configure HBase for secure operation. Refer to the section Secure Client Access to Apache HBase and complete all of the steps described there.

60.2. Prerequisites


60.3. Server-side Configuration for Simple User Access Operation

Add the following to the hbase-site.xml file on every server machine in the cluster:


For 0.94, add the following to the hbase-site.xml file on every server machine in the cluster:


A full shutdown and restart of HBase service is required when deploying these configuration changes.

60.4. Client-side Configuration for Simple User Access Operation

Add the following to the hbase-site.xml file on every client:


For 0.94, add the following to the hbase-site.xml file on every server machine in the cluster:


Be advised that if the hbase.security.authentication in the client- and server-side site files do not match, the client will not be able to communicate with the cluster.

60.4.1. Client-side Configuration for Simple User Access Operation - Thrift Gateway

The Thrift gateway user will need access. For example, to give the Thrift API user, thrift_server, administrative access, a command such as this one will suffice:

grant 'thrift_server', 'RWCA'

For more information about ACLs, please see the Access Control Labels (ACLs) section

The Thrift gateway will authenticate with HBase using the supplied credential. No authentication will be performed by the Thrift gateway itself. All client access via the Thrift gateway will use the Thrift gateway’s credential and have its privilege.

60.4.2. Client-side Configuration for Simple User Access Operation - REST Gateway

The REST gateway will authenticate with HBase using the supplied credential. No authentication will be performed by the REST gateway itself. All client access via the REST gateway will use the REST gateway’s credential and have its privilege.

The REST gateway user will need access. For example, to give the REST API user, rest_server, administrative access, a command such as this one will suffice:

grant 'rest_server', 'RWCA'

For more information about ACLs, please see the Access Control Labels (ACLs) section

It should be possible for clients to authenticate with the HBase cluster through the REST gateway in a pass-through manner via SPNEGO HTTP authentication. This is future work.

61. Securing Access to HDFS and ZooKeeper

Secure HBase requires secure ZooKeeper and HDFS so that users cannot access and/or modify the metadata and data from under HBase. HBase uses HDFS (or configured file system) to keep its data files as well as write ahead logs (WALs) and other data. HBase uses ZooKeeper to store some metadata for operations (master address, table locks, recovery state, etc).

61.1. Securing ZooKeeper Data

ZooKeeper has a pluggable authentication mechanism to enable access from clients using different methods. ZooKeeper even allows authenticated and un-authenticated clients at the same time. The access to znodes can be restricted by providing Access Control Lists (ACLs) per znode. An ACL contains two components, the authentication method and the principal. ACLs are NOT enforced hierarchically. See ZooKeeper Programmers Guide for details.

HBase daemons authenticate to ZooKeeper via SASL and kerberos (See SASL Authentication with ZooKeeper). HBase sets up the znode ACLs so that only the HBase user and the configured hbase superuser (hbase.superuser) can access and modify the data. In cases where ZooKeeper is used for service discovery or sharing state with the client, the znodes created by HBase will also allow anyone (regardless of authentication) to read these znodes (clusterId, master address, meta location, etc), but only the HBase user can modify them.

61.2. Securing File System (HDFS) Data

All of the data under management is kept under the root directory in the file system (hbase.rootdir). Access to the data and WAL files in the filesystem should be restricted so that users cannot bypass the HBase layer, and peek at the underlying data files from the file system. HBase assumes the filesystem used (HDFS or other) enforces permissions hierarchically. If sufficient protection from the file system (both authorization and authentication) is not provided, HBase level authorization control (ACLs, visibility labels, etc) is meaningless since the user can always access the data from the file system.

HBase enforces the posix-like permissions 700 (rwx------) to its root directory. It means that only the HBase user can read or write the files in FS. The default setting can be changed by configuring hbase.rootdir.perms in hbase-site.xml. A restart of the active master is needed so that it changes the used permissions. For versions before 1.2.0, you can check whether HBASE-13780 is committed, and if not, you can manually set the permissions for the root directory if needed. Using HDFS, the command would be:

sudo -u hdfs hadoop fs -chmod 700 /hbase

You should change /hbase if you are using a different hbase.rootdir.

In secure mode, SecureBulkLoadEndpoint should be configured and used for properly handing of users files created from MR jobs to the HBase daemons and HBase user. The staging directory in the distributed file system used for bulk load (hbase.bulkload.staging.dir, defaults to /tmp/hbase-staging) should have (mode 711, or rwx—​x—​x) so that users can access the staging directory created under that parent directory, but cannot do any other operation. See Secure Bulk Load for how to configure SecureBulkLoadEndPoint.

62. Securing Access To Your Data

After you have configured secure authentication between HBase client and server processes and gateways, you need to consider the security of your data itself. HBase provides several strategies for securing your data:

  • Role-based Access Control (RBAC) controls which users or groups can read and write to a given HBase resource or execute a coprocessor endpoint, using the familiar paradigm of roles.

  • Visibility Labels which allow you to label cells and control access to labelled cells, to further restrict who can read or write to certain subsets of your data. Visibility labels are stored as tags. See hbase.tags for more information.

  • Transparent encryption of data at rest on the underlying filesystem, both in HFiles and in the WAL. This protects your data at rest from an attacker who has access to the underlying filesystem, without the need to change the implementation of the client. It can also protect against data leakage from improperly disposed disks, which can be important for legal and regulatory compliance.

Server-side configuration, administration, and implementation details of each of these features are discussed below, along with any performance trade-offs. An example security configuration is given at the end, to show these features all used together, as they might be in a real-world scenario.

All aspects of security in HBase are in active development and evolving rapidly. Any strategy you employ for security of your data should be thoroughly tested. In addition, some of these features are still in the experimental stage of development. To take advantage of many of these features, you must be running HBase 0.98+ and using the HFile v3 file format.
Protecting Sensitive Files

Several procedures in this section require you to copy files between cluster nodes. When copying keys, configuration files, or other files containing sensitive strings, use a secure method, such as ssh, to avoid leaking sensitive data.

Procedure: Basic Server-Side Configuration
  1. Enable HFile v3, by setting hfile.format.version to 3 in hbase-site.xml. This is the default for HBase 1.0 and newer.

  2. Enable SASL and Kerberos authentication for RPC and ZooKeeper, as described in security.prerequisites and SASL Authentication with ZooKeeper.

62.1. Tags

Tags are a feature of HFile v3. A tag is a piece of metadata which is part of a cell, separate from the key, value, and version. Tags are an implementation detail which provides a foundation for other security-related features such as cell-level ACLs and visibility labels. Tags are stored in the HFiles themselves. It is possible that in the future, tags will be used to implement other HBase features. You don’t need to know a lot about tags in order to use the security features they enable.

62.1.1. Implementation Details

Every cell can have zero or more tags. Every tag has a type and the actual tag byte array.

Just as row keys, column families, qualifiers and values can be encoded (see data.block.encoding.types), tags can also be encoded as well. You can enable or disable tag encoding at the level of the column family, and it is enabled by default. Use the HColumnDescriptor#setCompressionTags(boolean compressTags) method to manage encoding settings on a column family. You also need to enable the DataBlockEncoder for the column family, for encoding of tags to take effect.

You can enable compression of each tag in the WAL, if WAL compression is also enabled, by setting the value of hbase.regionserver.wal.tags.enablecompression to true in hbase-site.xml. Tag compression uses dictionary encoding.

Tag compression is not supported when using WAL encryption.

62.2. Access Control Labels (ACLs)

62.2.1. How It Works

ACLs in HBase are based upon a user’s membership in or exclusion from groups, and a given group’s permissions to access a given resource. ACLs are implemented as a coprocessor called AccessController.

HBase does not maintain a private group mapping, but relies on a Hadoop group mapper, which maps between entities in a directory such as LDAP or Active Directory, and HBase users. Any supported Hadoop group mapper will work. Users are then granted specific permissions (Read, Write, Execute, Create, Admin) against resources (global, namespaces, tables, cells, or endpoints).

With Kerberos and Access Control enabled, client access to HBase is authenticated and user data is private unless access has been explicitly granted.

HBase has a simpler security model than relational databases, especially in terms of client operations. No distinction is made between an insert (new record) and update (of existing record), for example, as both collapse down into a Put.

Understanding Access Levels

HBase access levels are granted independently of each other and allow for different types of operations at a given scope.

  • Read (R) - can read data at the given scope

  • Write (W) - can write data at the given scope

  • Execute (X) - can execute coprocessor endpoints at the given scope

  • Create (C) - can create tables or drop tables (even those they did not create) at the given scope

  • Admin (A) - can perform cluster operations such as balancing the cluster or assigning regions at the given scope

The possible scopes are:

  • Superuser - superusers can perform any operation available in HBase, to any resource. The user who runs HBase on your cluster is a superuser, as are any principals assigned to the configuration property hbase.superuser in hbase-site.xml on the HMaster.

  • Global - permissions granted at global scope allow the admin to operate on all tables of the cluster.

  • Namespace - permissions granted at namespace scope apply to all tables within a given namespace.

  • Table - permissions granted at table scope apply to data or metadata within a given table.

  • ColumnFamily - permissions granted at ColumnFamily scope apply to cells within that ColumnFamily.

  • Cell - permissions granted at cell scope apply to that exact cell coordinate (key, value, timestamp). This allows for policy evolution along with data.

    To change an ACL on a specific cell, write an updated cell with new ACL to the precise coordinates of the original.

    If you have a multi-versioned schema and want to update ACLs on all visible versions, you need to write new cells for all visible versions. The application has complete control over policy evolution.

    The exception to the above rule is append and increment processing. Appends and increments can carry an ACL in the operation. If one is included in the operation, then it will be applied to the result of the append or increment. Otherwise, the ACL of the existing cell you are appending to or incrementing is preserved.

The combination of access levels and scopes creates a matrix of possible access levels that can be granted to a user. In a production environment, it is useful to think of access levels in terms of what is needed to do a specific job. The following list describes appropriate access levels for some common types of HBase users. It is important not to grant more access than is required for a given user to perform their required tasks.

  • Superusers - In a production system, only the HBase user should have superuser access. In a development environment, an administrator may need superuser access in order to quickly control and manage the cluster. However, this type of administrator should usually be a Global Admin rather than a superuser.

  • Global Admins - A global admin can perform tasks and access every table in HBase. In a typical production environment, an admin should not have Read or Write permissions to data within tables.

  • A global admin with Admin permissions can perform cluster-wide operations on the cluster, such as balancing, assigning or unassigning regions, or calling an explicit major compaction. This is an operations role.

  • A global admin with Create permissions can create or drop any table within HBase. This is more of a DBA-type role.

    In a production environment, it is likely that different users will have only one of Admin and Create permissions.

    In the current implementation, a Global Admin with Admin permission can grant himself Read and Write permissions on a table and gain access to that table’s data. For this reason, only grant Global Admin permissions to trusted user who actually need them.

    Also be aware that a Global Admin with Create permission can perform a Put operation on the ACL table, simulating a grant or revoke and circumventing the authorization check for Global Admin permissions.

    Due to these issues, be cautious with granting Global Admin privileges.

  • Namespace Admins - a namespace admin with Create permissions can create or drop tables within that namespace, and take and restore snapshots. A namespace admin with Admin permissions can perform operations such as splits or major compactions on tables within that namespace.

  • Table Admins - A table admin can perform administrative operations only on that table. A table admin with Create permissions can create snapshots from that table or restore that table from a snapshot. A table admin with Admin permissions can perform operations such as splits or major compactions on that table.

  • Users - Users can read or write data, or both. Users can also execute coprocessor endpoints, if given Executable permissions.

Table 7. Real-World Example of Access Levels
Job Title Scope Permissions Description

Senior Administrator


Access, Create

Manages the cluster and gives access to Junior Administrators.

Junior Administrator



Creates tables and gives access to Table Administrators.

Table Administrator



Maintains a table from an operations point of view.

Data Analyst



Creates reports from HBase data.

Web Application


Read, Write

Puts data into HBase and uses HBase data to perform operations.

ACL Matrix

For more details on how ACLs map to specific HBase operations and tasks, see appendix acl matrix.

Implementation Details

Cell-level ACLs are implemented using tags (see Tags). In order to use cell-level ACLs, you must be using HFile v3 and HBase 0.98 or newer.

  1. Files created by HBase are owned by the operating system user running the HBase process. To interact with HBase files, you should use the API or bulk load facility.

  2. HBase does not model "roles" internally in HBase. Instead, group names can be granted permissions. This allows external modeling of roles via group membership. Groups are created and manipulated externally to HBase, via the Hadoop group mapping service.

Server-Side Configuration
  1. As a prerequisite, perform the steps in Procedure: Basic Server-Side Configuration.

  2. Install and configure the AccessController coprocessor, by setting the following properties in hbase-site.xml. These properties take a list of classes.

    If you use the AccessController along with the VisibilityController, the AccessController must come first in the list, because with both components active, the VisibilityController will delegate access control on its system tables to the AccessController. For an example of using both together, see Security Configuration Example.
      <value>org.apache.hadoop.hbase.security.access.AccessController, org.apache.hadoop.hbase.security.token.TokenProvider</value>

    Optionally, you can enable transport security, by setting hbase.rpc.protection to privacy. This requires HBase 0.98.4 or newer.

  3. Set up the Hadoop group mapper in the Hadoop namenode’s core-site.xml. This is a Hadoop file, not an HBase file. Customize it to your site’s needs. Following is an example.

  4. Optionally, enable the early-out evaluation strategy. Prior to HBase 0.98.0, if a user was not granted access to a column family, or at least a column qualifier, an AccessDeniedException would be thrown. HBase 0.98.0 removed this exception in order to allow cell-level exceptional grants. To restore the old behavior in HBase 0.98.0-0.98.6, set hbase.security.access.early_out to true in hbase-site.xml. In HBase 0.98.6, the default has been returned to true.

  5. Distribute your configuration and restart your cluster for changes to take effect.

  6. To test your configuration, log into HBase Shell as a given user and use the whoami command to report the groups your user is part of. In this example, the user is reported as being a member of the services group.

    hbase> whoami
    service (auth:KERBEROS)
        groups: services

Administration tasks can be performed from HBase Shell or via an API.

API Examples

Many of the API examples below are taken from source files hbase-server/src/test/java/org/apache/hadoop/hbase/security/access/TestAccessController.java and hbase-server/src/test/java/org/apache/hadoop/hbase/security/access/SecureTestUtil.java.

Neither the examples, nor the source files they are taken from, are part of the public HBase API, and are provided for illustration only. Refer to the official API for usage instructions.

  1. User and Group Administration

    Users and groups are maintained external to HBase, in your directory.

  2. Granting Access To A Namespace, Table, Column Family, or Cell

    There are a few different types of syntax for grant statements. The first, and most familiar, is as follows, with the table and column family being optional:

    grant 'user', 'RWXCA', 'TABLE', 'CF', 'CQ'

    Groups and users are granted access in the same way, but groups are prefixed with an @ symbol. In the same way, tables and namespaces are specified in the same way, but namespaces are prefixed with an @ symbol.

    It is also possible to grant multiple permissions against the same resource in a single statement, as in this example. The first sub-clause maps users to ACLs and the second sub-clause specifies the resource.

    HBase Shell support for granting and revoking access at the cell level is for testing and verification support, and should not be employed for production use because it won’t apply the permissions to cells that don’t exist yet. The correct way to apply cell level permissions is to do so in the application code when storing the values.
    ACL Granularity and Evaluation Order

    ACLs are evaluated from least granular to most granular, and when an ACL is reached that grants permission, evaluation stops. This means that cell ACLs do not override ACLs at less granularity.

    Example 20. HBase Shell
    • Global:

      hbase> grant '@admins', 'RWXCA'
    • Namespace:

      hbase> grant 'service', 'RWXCA', '@test-NS'
    • Table:

      hbase> grant 'service', 'RWXCA', 'user'
    • Column Family:

      hbase> grant '@developers', 'RW', 'user', 'i'
    • Column Qualifier:

      hbase> grant 'service, 'RW', 'user', 'i', 'foo'
    • Cell:

      The syntax for granting cell ACLs uses the following syntax:

      grant <table>, \
        { '<user-or-group>' => \
          '<permissions>', ... }, \
        { <scanner-specification> }
    • <user-or-group> is the user or group name, prefixed with @ in the case of a group.

    • <permissions> is a string containing any or all of "RWXCA", though only R and W are meaningful at cell scope.

    • <scanner-specification> is the scanner specification syntax and conventions used by the 'scan' shell command. For some examples of scanner specifications, issue the following HBase Shell command.

      hbase> help "scan"

      This example grants read access to the 'testuser' user and read/write access to the 'developers' group, on cells in the 'pii' column which match the filter.

      hbase> grant 'user', \
        { '@developers' => 'RW', 'testuser' => 'R' }, \
        { COLUMNS => 'pii', FILTER => "(PrefixFilter ('test'))" }

      The shell will run a scanner with the given criteria, rewrite the found cells with new ACLs, and store them back to their exact coordinates.

    Example 21. API

    The following example shows how to grant access at the table level.

    public static void grantOnTable(final HBaseTestingUtility util, final String user,
        final TableName table, final byte[] family, final byte[] qualifier,
        final Permission.Action... actions) throws Exception {
      SecureTestUtil.updateACLs(util, new Callable<Void>() {
        public Void call() throws Exception {
          Configuration conf = HBaseConfiguration.create();
          Connection connection = ConnectionFactory.createConnection(conf);
          try (Connection connection = ConnectionFactory.createConnection(conf)) {
            try (Table table = connection.getTable(TableName.valueOf(tablename)) {
              try {
                BlockingRpcChannel service = acl.coprocessorService(HConstants.EMPTY_START_ROW);
                AccessControlService.BlockingInterface protocol =
                ProtobufUtil.grant(protocol, user, table, family, qualifier, actions);
              } finally {
              return null;

    To grant permissions at the cell level, you can use the Mutation.setACL method:

    Mutation.setACL(String user, Permission perms)
    Mutation.setACL(Map<String, Permission> perms)

    Specifically, this example provides read permission to a user called user1 on any cells contained in a particular Put operation:

    put.setACL(user1, new Permission(Permission.Action.READ))
  3. Revoking Access Control From a Namespace, Table, Column Family, or Cell

    The revoke command and API are twins of the grant command and API, and the syntax is exactly the same. The only exception is that you cannot revoke permissions at the cell level. You can only revoke access that has previously been granted, and a revoke statement is not the same thing as explicit denial to a resource.

    HBase Shell support for granting and revoking access is for testing and verification support, and should not be employed for production use because it won’t apply the permissions to cells that don’t exist yet. The correct way to apply cell-level permissions is to do so in the application code when storing the values.
    Example 22. Revoking Access To a Table
    public static void revokeFromTable(final HBaseTestingUtility util, final String user,
        final TableName table, final byte[] family, final byte[] qualifier,
        final Permission.Action... actions) throws Exception {
      SecureTestUtil.updateACLs(util, new Callable<Void>() {
        public Void call() throws Exception {
          Configuration conf = HBaseConfiguration.create();
          Connection connection = ConnectionFactory.createConnection(conf);
          Table acl = connection.getTable(util.getConfiguration(), AccessControlLists.ACL_TABLE_NAME);
          try {
            BlockingRpcChannel service = acl.coprocessorService(HConstants.EMPTY_START_ROW);
            AccessControlService.BlockingInterface protocol =
            ProtobufUtil.revoke(protocol, user, table, family, qualifier, actions);
          } finally {
          return null;
  4. Showing a User’s Effective Permissions

    Example 23. HBase Shell
    hbase> user_permission 'user'
    hbase> user_permission '.*'
    hbase> user_permission JAVA_REGEX
Example 24. API
public static void verifyAllowed(User user, AccessTestAction action, int count) throws Exception {
  try {
    Object obj = user.runAs(action);
    if (obj != null && obj instanceof List&lt;?&gt;) {
      List&lt;?&gt; results = (List&lt;?&gt;) obj;
      if (results != null && results.isEmpty()) {
        fail("Empty non null results from action for user '" ` user.getShortName() ` "'");
      assertEquals(count, results.size());
  } catch (AccessDeniedException ade) {
    fail("Expected action to pass for user '" ` user.getShortName() ` "' but was denied");

62.3. Visibility Labels

Visibility labels control can be used to only permit users or principals associated with a given label to read or access cells with that label. For instance, you might label a cell top-secret, and only grant access to that label to the managers group. Visibility labels are implemented using Tags, which are a feature of HFile v3, and allow you to store metadata on a per-cell basis. A label is a string, and labels can be combined into expressions by using logical operators (&, |, or !), and using parentheses for grouping. HBase does not do any kind of validation of expressions beyond basic well-formedness. Visibility labels have no meaning on their own, and may be used to denote sensitivity level, privilege level, or any other arbitrary semantic meaning.

If a user’s labels do not match a cell’s label or expression, the user is denied access to the cell.

In HBase 0.98.6 and newer, UTF-8 encoding is supported for visibility labels and expressions. When creating labels using the addLabels(conf, labels) method provided by the org.apache.hadoop.hbase.security.visibility.VisibilityClient class and passing labels in Authorizations via Scan or Get, labels can contain UTF-8 characters, as well as the logical operators normally used in visibility labels, with normal Java notations, without needing any escaping method. However, when you pass a CellVisibility expression via a Mutation, you must enclose the expression with the CellVisibility.quote() method if you use UTF-8 characters or logical operators. See TestExpressionParser and the source file hbase-client/src/test/java/org/apache/hadoop/hbase/client/TestScan.java.

A user adds visibility expressions to a cell during a Put operation. In the default configuration, the user does not need to have access to a label in order to label cells with it. This behavior is controlled by the configuration option hbase.security.visibility.mutations.checkauths. If you set this option to true, the labels the user is modifying as part of the mutation must be associated with the user, or the mutation will fail. Whether a user is authorized to read a labelled cell is determined during a Get or Scan, and results which the user is not allowed to read are filtered out. This incurs the same I/O penalty as if the results were returned, but reduces load on the network.

Visibility labels can also be specified during Delete operations. For details about visibility labels and Deletes, see HBASE-10885.

The user’s effective label set is built in the RPC context when a request is first received by the RegionServer. The way that users are associated with labels is pluggable. The default plugin passes through labels specified in Authorizations added to the Get or Scan and checks those against the calling user’s authenticated labels list. When the client passes labels for which the user is not authenticated, the default plugin drops them. You can pass a subset of user authenticated labels via the Get#setAuthorizations(Authorizations(String,…​)) and Scan#setAuthorizations(Authorizations(String,…​)); methods.

Groups can be granted visibility labels the same way as users. Groups are prefixed with an @ symbol. When checking visibility labels of a user, the server will include the visibility labels of the groups of which the user is a member, together with the user’s own labels. When the visibility labels are retrieved using API VisibilityClient#getAuths or Shell command get_auths for a user, we will return labels added specifically for that user alone, not the group level labels.

Visibility label access checking is performed by the VisibilityController coprocessor. You can use interface VisibilityLabelService to provide a custom implementation and/or control the way that visibility labels are stored with cells. See the source file hbase-server/src/test/java/org/apache/hadoop/hbase/security/visibility/TestVisibilityLabelsWithCustomVisLabService.java for one example.

Visibility labels can be used in conjunction with ACLs.

The labels have to be explicitly defined before they can be used in visibility labels. See below for an example of how this can be done.
There is currently no way to determine which labels have been applied to a cell. See HBASE-12470 for details.
Visibility labels are not currently applied for superusers.
Table 8. Examples of Visibility Expressions
Expression Interpretation

Allow access to users associated with the fulltime label.


Allow access to users not associated with the public label.

( secret | topsecret ) & !probationary

Allow access to users associated with either the secret or topsecret label and not associated with the probationary label.

62.3.1. Server-Side Configuration

  1. As a prerequisite, perform the steps in Procedure: Basic Server-Side Configuration.

  2. Install and configure the VisibilityController coprocessor by setting the following properties in hbase-site.xml. These properties take a list of class names.

    If you use the AccessController and VisibilityController coprocessors together, the AccessController must come first in the list, because with both components active, the VisibilityController will delegate access control on its system tables to the AccessController.
  3. Adjust Configuration

    By default, users can label cells with any label, including labels they are not associated with, which means that a user can Put data that he cannot read. For example, a user could label a cell with the (hypothetical) 'topsecret' label even if the user is not associated with that label. If you only want users to be able to label cells with labels they are associated with, set hbase.security.visibility.mutations.checkauths to true. In that case, the mutation will fail if it makes use of labels the user is not associated with.

  4. Distribute your configuration and restart your cluster for changes to take effect.

62.3.2. Administration

Administration tasks can be performed using the HBase Shell or the Java API. For defining the list of visibility labels and associating labels with users, the HBase Shell is probably simpler.

API Examples

Many of the Java API examples in this section are taken from the source file hbase-server/src/test/java/org/apache/hadoop/hbase/security/visibility/TestVisibilityLabels.java. Refer to that file or the API documentation for more context.

Neither these examples, nor the source file they were taken from, are part of the public HBase API, and are provided for illustration only. Refer to the official API for usage instructions.

  1. Define the List of Visibility Labels

    Example 25. HBase Shell
    hbase> add_labels [ 'admin', 'service', 'developer', 'test' ]
    Example 26. Java API
    public static void addLabels() throws Exception {
      PrivilegedExceptionAction<VisibilityLabelsResponse> action = new PrivilegedExceptionAction<VisibilityLabelsResponse>() {
        public VisibilityLabelsResponse run() throws Exception {
              UNICODE_VIS_TAG, UC1, UC2 };
          try {
            VisibilityClient.addLabels(conf, labels);
          } catch (Throwable t) {
            throw new IOException(t);
          return null;
  2. Associate Labels with Users

    Example 27. HBase Shell
    hbase> set_auths 'service', [ 'service' ]
    hbase> set_auths 'testuser', [ 'test' ]
    hbase> set_auths 'qa', [ 'test', 'developer' ]
    hbase> set_auths '@qagroup', [ 'test' ]
    Example 28. Java API
    public void testSetAndGetUserAuths() throws Throwable {
      final String user = "user1";
      PrivilegedExceptionAction<Void> action = new PrivilegedExceptionAction<Void>() {
        public Void run() throws Exception {
          String[] auths = { SECRET, CONFIDENTIAL };
          try {
            VisibilityClient.setAuths(conf, auths, user);
          } catch (Throwable e) {
          return null;
  3. Clear Labels From Users

    Example 29. HBase Shell
    hbase> clear_auths 'service', [ 'service' ]
    hbase> clear_auths 'testuser', [ 'test' ]
    hbase> clear_auths 'qa', [ 'test', 'developer' ]
    hbase> clear_auths '@qagroup', [ 'test', 'developer' ]
    Example 30. Java API
    auths = new String[] { SECRET, PUBLIC, CONFIDENTIAL };
    VisibilityLabelsResponse response = null;
    try {
      response = VisibilityClient.clearAuths(conf, auths, user);
    } catch (Throwable e) {
      fail("Should not have failed");
  4. Apply a Label or Expression to a Cell

    The label is only applied when data is written. The label is associated with a given version of the cell.

    Example 31. HBase Shell
    hbase> set_visibility 'user', 'admin|service|developer', { COLUMNS => 'i' }
    hbase> set_visibility 'user', 'admin|service', { COLUMNS => 'pii' }
    hbase> set_visibility 'user', 'test', { COLUMNS => [ 'i', 'pii' ], FILTER => "(PrefixFilter ('test'))" }
    HBase Shell support for applying labels or permissions to cells is for testing and verification support, and should not be employed for production use because it won’t apply the labels to cells that don’t exist yet. The correct way to apply cell level labels is to do so in the application code when storing the values.
    Example 32. Java API
    static Table createTableAndWriteDataWithLabels(TableName tableName, String... labelExps)
        throws Exception {
      Configuration conf = HBaseConfiguration.create();
      Connection connection = ConnectionFactory.createConnection(conf);
      Table table = NULL;
      try {
        table = TEST_UTIL.createTable(tableName, fam);
        int i = 1;
        List<Put> puts = new ArrayList<Put>();
        for (String labelExp : labelExps) {
          Put put = new Put(Bytes.toBytes("row" + i));
          put.add(fam, qual, HConstants.LATEST_TIMESTAMP, value);
          put.setCellVisibility(new CellVisibility(labelExp));
      } finally {
        if (table != null) {

62.3.3. Reading Cells with Labels

When you issue a Scan or Get, HBase uses your default set of authorizations to filter out cells that you do not have access to. A superuser can set the default set of authorizations for a given user by using the set_auths HBase Shell command or the ],%20java.lang.String)[VisibilityClient.setAuths() method.

You can specify a different authorization during the Scan or Get, by passing the AUTHORIZATIONS option in HBase Shell, or the setAuthorizations() method if you use the API. This authorization will be combined with your default set as an additional filter. It will further filter your results, rather than giving you additional authorization.

Example 33. HBase Shell
hbase> get_auths 'myUser'
hbase> scan 'table1', AUTHORIZATIONS => ['private']
Example 34. Java API
public Void run() throws Exception {
  String[] auths1 = { SECRET, CONFIDENTIAL };
  GetAuthsResponse authsResponse = null;
  try {
    VisibilityClient.setAuths(conf, auths1, user);
    try {
      authsResponse = VisibilityClient.getAuths(conf, user);
    } catch (Throwable e) {
      fail("Should not have failed");
  } catch (Throwable e) {
  List<String> authsList = new ArrayList<String>();
  for (ByteString authBS : authsResponse.getAuthList()) {
  assertEquals(2, authsList.size());
  return null;

62.3.4. Implementing Your Own Visibility Label Algorithm

Interpreting the labels authenticated for a given get/scan request is a pluggable algorithm.

You can specify a custom plugin or plugins by using the property hbase.regionserver.scan.visibility.label.generator.class. The output for the first ScanLabelGenerator will be the input for the next one, until the end of the list.

The default implementation, which was implemented in HBASE-12466, loads two plugins, FeedUserAuthScanLabelGenerator and DefinedSetFilterScanLabelGenerator. See Reading Cells with Labels.

62.3.5. Replicating Visibility Tags as Strings

As mentioned in the above sections, the interface VisibilityLabelService could be used to implement a different way of storing the visibility expressions in the cells. Clusters with replication enabled also must replicate the visibility expressions to the peer cluster. If DefaultVisibilityLabelServiceImpl is used as the implementation for VisibilityLabelService, all the visibility expression are converted to the corresponding expression based on the ordinals for each visibility label stored in the labels table. During replication, visible cells are also replicated with the ordinal-based expression intact. The peer cluster may not have the same labels table with the same ordinal mapping for the visibility labels. In that case, replicating the ordinals makes no sense. It would be better if the replication occurred with the visibility expressions transmitted as strings. To replicate the visibility expression as strings to the peer cluster, create a RegionServerObserver configuration which works based on the implementation of the VisibilityLabelService interface. The configuration below enables replication of visibility expressions to peer clusters as strings. See HBASE-11639 for more details.


62.4. Transparent Encryption of Data At Rest

HBase provides a mechanism for protecting your data at rest, in HFiles and the WAL, which reside within HDFS or another distributed filesystem. A two-tier architecture is used for flexible and non-intrusive key rotation. "Transparent" means that no implementation changes are needed on the client side. When data is written, it is encrypted. When it is read, it is decrypted on demand.

62.4.1. How It Works

The administrator provisions a master key for the cluster, which is stored in a key provider accessible to every trusted HBase process, including the HMaster, RegionServers, and clients (such as HBase Shell) on administrative workstations. The default key provider is integrated with the Java KeyStore API and any key management systems with support for it. Other custom key provider implementations are possible. The key retrieval mechanism is configured in the hbase-site.xml configuration file. The master key may be stored on the cluster servers, protected by a secure KeyStore file, or on an external keyserver, or in a hardware security module. This master key is resolved as needed by HBase processes through the configured key provider.

Next, encryption use can be specified in the schema, per column family, by creating or modifying a column descriptor to include two additional attributes: the name of the encryption algorithm to use (currently only "AES" is supported), and optionally, a data key wrapped (encrypted) with the cluster master key. If a data key is not explicitly configured for a ColumnFamily, HBase will create a random data key per HFile. This provides an incremental improvement in security over the alternative. Unless you need to supply an explicit data key, such as in a case where you are generating encrypted HFiles for bulk import with a given data key, only specify the encryption algorithm in the ColumnFamily schema metadata and let HBase create data keys on demand. Per Column Family keys facilitate low impact incremental key rotation and reduce the scope of any external leak of key material. The wrapped data key is stored in the ColumnFamily schema metadata, and in each HFile for the Column Family, encrypted with the cluster master key. After the Column Family is configured for encryption, any new HFiles will be written encrypted. To ensure encryption of all HFiles, trigger a major compaction after enabling this feature.

When the HFile is opened, the data key is extracted from the HFile, decrypted with the cluster master key, and used for decryption of the remainder of the HFile. The HFile will be unreadable if the master key is not available. If a remote user somehow acquires access to the HFile data because of some lapse in HDFS permissions, or from inappropriately discarded media, it will not be possible to decrypt either the data key or the file data.

It is also possible to encrypt the WAL. Even though WALs are transient, it is necessary to encrypt the WALEdits to avoid circumventing HFile protections for encrypted column families, in the event that the underlying filesystem is compromised. When WAL encryption is enabled, all WALs are encrypted, regardless of whether the relevant HFiles are encrypted.

62.4.2. Server-Side Configuration

This procedure assumes you are using the default Java keystore implementation. If you are using a custom implementation, check its documentation and adjust accordingly.

  1. Create a secret key of appropriate length for AES encryption, using the keytool utility.

    $ keytool -keystore /path/to/hbase/conf/hbase.jks \
      -storetype jceks -storepass **** \
      -genseckey -keyalg AES -keysize 128 \
      -alias <alias>

    Replace **** with the password for the keystore file and <alias> with the username of the HBase service account, or an arbitrary string. If you use an arbitrary string, you will need to configure HBase to use it, and that is covered below. Specify a keysize that is appropriate. Do not specify a separate password for the key, but press Return when prompted.

  2. Set appropriate permissions on the keyfile and distribute it to all the HBase servers.

    The previous command created a file called hbase.jks in the HBase conf/ directory. Set the permissions and ownership on this file such that only the HBase service account user can read the file, and securely distribute the key to all HBase servers.

  3. Configure the HBase daemons.

    Set the following properties in hbase-site.xml on the region servers, to configure HBase daemons to use a key provider backed by the KeyStore file or retrieving the cluster master key. In the example below, replace **** with the password.


    By default, the HBase service account name will be used to resolve the cluster master key. However, you can store it with an arbitrary alias (in the keytool command). In that case, set the following property to the alias you used.


    You also need to be sure your HFiles use HFile v3, in order to use transparent encryption. This is the default configuration for HBase 1.0 onward. For previous versions, set the following property in your hbase-site.xml file.


    Optionally, you can use a different cipher provider, either a Java Cryptography Encryption (JCE) algorithm provider or a custom HBase cipher implementation.

    • JCE:

      • Install a signed JCE provider (supporting AES/CTR/NoPadding mode with 128 bit keys)

      • Add it with highest preference to the JCE site configuration file $JAVA_HOME/lib/security/java.security.

      • Update hbase.crypto.algorithm.aes.provider and hbase.crypto.algorithm.rng.provider options in hbase-site.xml.

    • Custom HBase Cipher:

      • Implement org.apache.hadoop.hbase.io.crypto.CipherProvider.

      • Add the implementation to the server classpath.

      • Update hbase.crypto.cipherprovider in hbase-site.xml.

  4. Configure WAL encryption.

    Configure WAL encryption in every RegionServer’s hbase-site.xml, by setting the following properties. You can include these in the HMaster’s hbase-site.xml as well, but the HMaster does not have a WAL and will not use them.

  5. Configure permissions on the hbase-site.xml file.

    Because the keystore password is stored in the hbase-site.xml, you need to ensure that only the HBase user can read the hbase-site.xml file, using file ownership and permissions.

  6. Restart your cluster.

    Distribute the new configuration file to all nodes and restart your cluster.

62.4.3. Administration

Administrative tasks can be performed in HBase Shell or the Java API.

Java API

Java API examples in this section are taken from the source file hbase-server/src/test/java/org/apache/hadoop/hbase/util/TestHBaseFsckEncryption.java. .

Neither these examples, nor the source files they are taken from, are part of the public HBase API, and are provided for illustration only. Refer to the official API for usage instructions.

Enable Encryption on a Column Family

To enable encryption on a column family, you can either use HBase Shell or the Java API. After enabling encryption, trigger a major compaction. When the major compaction completes, the HFiles will be encrypted.

Rotate the Data Key

To rotate the data key, first change the ColumnFamily key in the column descriptor, then trigger a major compaction. When compaction is complete, all HFiles will be re-encrypted using the new data key. Until the compaction completes, the old HFiles will still be readable using the old key.

Switching Between Using a Random Data Key and Specifying A Key

If you configured a column family to use a specific key and you want to return to the default behavior of using a randomly-generated key for that column family, use the Java API to alter the HColumnDescriptor so that no value is sent with the key ENCRYPTION_KEY.

Rotate the Master Key

To rotate the master key, first generate and distribute the new key. Then update the KeyStore to contain a new master key, and keep the old master key in the KeyStore using a different alias. Next, configure fallback to the old master key in the hbase-site.xml file.

62.5. Secure Bulk Load

Bulk loading in secure mode is a bit more involved than normal setup, since the client has to transfer the ownership of the files generated from the MapReduce job to HBase. Secure bulk loading is implemented by a coprocessor, named SecureBulkLoadEndpoint, which uses a staging directory configured by the configuration property hbase.bulkload.staging.dir, which defaults to /tmp/hbase-staging/.

Secure Bulk Load Algorithm
  • One time only, create a staging directory which is world-traversable and owned by the user which runs HBase (mode 711, or rwx—​x—​x). A listing of this directory will look similar to the following:

    $ ls -ld /tmp/hbase-staging
    drwx--x--x  2 hbase  hbase  68  3 Sep 14:54 /tmp/hbase-staging
  • A user writes out data to a secure output directory owned by that user. For example, /user/foo/data.

  • Internally, HBase creates a secret staging directory which is globally readable/writable (-rwxrwxrwx, 777). For example, /tmp/hbase-staging/averylongandrandomdirectoryname. The name and location of this directory is not exposed to the user. HBase manages creation and deletion of this directory.

  • The user makes the data world-readable and world-writable, moves it into the random staging directory, then calls the SecureBulkLoadClient#bulkLoadHFiles method.

The strength of the security lies in the length and randomness of the secret directory.

To enable secure bulk load, add the following properties to hbase-site.xml.


63. Security Configuration Example

This configuration example includes support for HFile v3, ACLs, Visibility Labels, and transparent encryption of data at rest and the WAL. All options have been discussed separately in the sections above.

Example 35. Example Security Settings in hbase-site.xml
<!-- HFile v3 Support -->
<!-- HBase Superuser -->
  <value>hbase, admin</value>
<!-- Coprocessors for ACLs and Visibility Tags -->
<!-- Executable ACL for Coprocessor Endpoints -->
<!-- Whether a user needs authorization for a visibility tag to set it on a cell -->
<!-- Secure RPC Transport -->
 <!-- Transparent Encryption -->
<!-- WAL Encryption -->
<!-- For key rotation -->
<!-- Secure Bulk Load -->
Example 36. Example Group Mapper in Hadoop core-site.xml

Adjust these settings to suit your environment.

  <value>****</value> <!-- Replace with the actual password -->


64. Overview

64.1. NoSQL?

HBase is a type of "NoSQL" database. "NoSQL" is a general term meaning that the database isn’t an RDBMS which supports SQL as its primary access language, but there are many types of NoSQL databases: BerkeleyDB is an example of a local NoSQL database, whereas HBase is very much a distributed database. Technically speaking, HBase is really more a "Data Store" than "Data Base" because it lacks many of the features you find in an RDBMS, such as typed columns, secondary indexes, triggers, and advanced query languages, etc.

However, HBase has many features which supports both linear and modular scaling. HBase clusters expand by adding RegionServers that are hosted on commodity class servers. If a cluster expands from 10 to 20 RegionServers, for example, it doubles both in terms of storage and as well as processing capacity. An RDBMS can scale well, but only up to a point - specifically, the size of a single database server - and for the best performance requires specialized hardware and storage devices. HBase features of note are:

  • Strongly consistent reads/writes: HBase is not an "eventually consistent" DataStore. This makes it very suitable for tasks such as high-speed counter aggregation.

  • Automatic sharding: HBase tables are distributed on the cluster via regions, and regions are automatically split and re-distributed as your data grows.

  • Automatic RegionServer failover

  • Hadoop/HDFS Integration: HBase supports HDFS out of the box as its distributed file system.

  • MapReduce: HBase supports massively parallelized processing via MapReduce for using HBase as both source and sink.

  • Java Client API: HBase supports an easy to use Java API for programmatic access.

  • Thrift/REST API: HBase also supports Thrift and REST for non-Java front-ends.

  • Block Cache and Bloom Filters: HBase supports a Block Cache and Bloom Filters for high volume query optimization.

  • Operational Management: HBase provides build-in web-pages for operational insight as well as JMX metrics.

64.2. When Should I Use HBase?

HBase isn’t suitable for every problem.

First, make sure you have enough data. If you have hundreds of millions or billions of rows, then HBase is a good candidate. If you only have a few thousand/million rows, then using a traditional RDBMS might be a better choice due to the fact that all of your data might wind up on a single node (or two) and the rest of the cluster may be sitting idle.

Second, make sure you can live without all the extra features that an RDBMS provides (e.g., typed columns, secondary indexes, transactions, advanced query languages, etc.) An application built against an RDBMS cannot be "ported" to HBase by simply changing a JDBC driver, for example. Consider moving from an RDBMS to HBase as a complete redesign as opposed to a port.

Third, make sure you have enough hardware. Even HDFS doesn’t do well with anything less than 5 DataNodes (due to things such as HDFS block replication which has a default of 3), plus a NameNode.

HBase can run quite well stand-alone on a laptop - but this should be considered a development configuration only.

64.3. What Is The Difference Between HBase and Hadoop/HDFS?

HDFS is a distributed file system that is well suited for the storage of large files. Its documentation states that it is not, however, a general purpose file system, and does not provide fast individual record lookups in files. HBase, on the other hand, is built on top of HDFS and provides fast record lookups (and updates) for large tables. This can sometimes be a point of conceptual confusion. HBase internally puts your data in indexed "StoreFiles" that exist on HDFS for high-speed lookups. See the Data Model and the rest of this chapter for more information on how HBase achieves its goals.

65. Catalog Tables

The catalog table hbase:meta exists as an HBase table and is filtered out of the HBase shell’s list command, but is in fact a table just like any other.

65.1. -ROOT-

The -ROOT- table was removed in HBase 0.96.0. Information here should be considered historical.

The -ROOT- table kept track of the location of the .META table (the previous name for the table now called hbase:meta) prior to HBase 0.96. The -ROOT- table structure was as follows:

  • .META. region key (.META.,,1)

  • info:regioninfo (serialized HRegionInfo instance of hbase:meta)

  • info:server (server:port of the RegionServer holding hbase:meta)

  • info:serverstartcode (start-time of the RegionServer process holding hbase:meta)

65.2. hbase:meta

The hbase:meta table (previously called .META.) keeps a list of all regions in the system. The location of hbase:meta was previously tracked within the -ROOT- table, but is now stored in ZooKeeper.

The hbase:meta table structure is as follows:

  • Region key of the format ([table],[region start key],[region id])

  • info:regioninfo (serialized HRegionInfo instance for this region)

  • info:server (server:port of the RegionServer containing this region)

  • info:serverstartcode (start-time of the RegionServer process containing this region)

When a table is in the process of splitting, two other columns will be created, called info:splitA and info:splitB. These columns represent the two daughter regions. The values for these columns are also serialized HRegionInfo instances. After the region has been split, eventually this row will be deleted.

Note on HRegionInfo

The empty key is used to denote table start and table end. A region with an empty start key is the first region in a table. If a region has both an empty start and an empty end key, it is the only region in the table

In the (hopefully unlikely) event that programmatic processing of catalog metadata is required, see the Writables utility.

65.3. Startup Sequencing

First, the location of hbase:meta is looked up in ZooKeeper. Next, hbase:meta is updated with server and startcode values.

For information on region-RegionServer assignment, see Region-RegionServer Assignment.

66. Client

The HBase client finds the RegionServers that are serving the particular row range of interest. It does this by querying the hbase:meta table. See hbase:meta for details. After locating the required region(s), the client contacts the RegionServer serving that region, rather than going through the master, and issues the read or write request. This information is cached in the client so that subsequent requests need not go through the lookup process. Should a region be reassigned either by the master load balancer or because a RegionServer has died, the client will requery the catalog tables to determine the new location of the user region.

See Runtime Impact for more information about the impact of the Master on HBase Client communication.

Administrative functions are done via an instance of Admin

66.1. Cluster Connections

The API changed in HBase 1.0. For connection configuration information, see Client configuration and dependencies connecting to an HBase cluster.

66.1.1. API as of HBase 1.0.0

It’s been cleaned up and users are returned Interfaces to work against rather than particular types. In HBase 1.0, obtain a Connection object from ConnectionFactory and thereafter, get from it instances of Table, Admin, and RegionLocator on an as-need basis. When done, close the obtained instances. Finally, be sure to cleanup your Connection instance before exiting. Connections are heavyweight objects but thread-safe so you can create one for your application and keep the instance around. Table, Admin and RegionLocator instances are lightweight. Create as you go and then let go as soon as you are done by closing them. See the Client Package Javadoc Description for example usage of the new HBase 1.0 API.

66.1.2. API before HBase 1.0.0

Instances of HTable are the way to interact with an HBase cluster earlier than 1.0.0. Table instances are not thread-safe. Only one thread can use an instance of Table at any given time. When creating Table instances, it is advisable to use the same HBaseConfiguration instance. This will ensure sharing of ZooKeeper and socket instances to the RegionServers which is usually what you want. For example, this is preferred:

HBaseConfiguration conf = HBaseConfiguration.create();
HTable table1 = new HTable(conf, "myTable");
HTable table2 = new HTable(conf, "myTable");

as opposed to this:

HBaseConfiguration conf1 = HBaseConfiguration.create();
HTable table1 = new HTable(conf1, "myTable");
HBaseConfiguration conf2 = HBaseConfiguration.create();
HTable table2 = new HTable(conf2, "myTable");

For more information about how connections are handled in the HBase client, see ConnectionFactory.

Connection Pooling

For applications which require high-end multithreaded access (e.g., web-servers or application servers that may serve many application threads in a single JVM), you can pre-create a Connection, as shown in the following example:

Example 37. Pre-Creating a Connection
// Create a connection to the cluster.
Configuration conf = HBaseConfiguration.create();
try (Connection connection = ConnectionFactory.createConnection(conf)) {
  try (Table table = connection.getTable(TableName.valueOf(tablename)) {
    // use table as needed, the table returned is lightweight

Constructing HTableInterface implementation is very lightweight and resources are controlled.

HTablePool is Deprecated

Previous versions of this guide discussed HTablePool, which was deprecated in HBase 0.94, 0.95, and 0.96, and removed in 0.98.1, by HBASE-6500, or HConnection, which is deprecated in HBase 1.0 by Connection. Please use Connection instead.

66.2. WriteBuffer and Batch Methods

In HBase 1.0 and later, HTable is deprecated in favor of Table. Table does not use autoflush. To do buffered writes, use the BufferedMutator class.

Before a Table or HTable instance is discarded, invoke either close() or flushCommits(), so `Put`s will not be lost.

For additional information on write durability, review the ACID semantics page.

For fine-grained control of batching of Puts or Deletes, see the batch methods on Table.

66.3. External Clients

Information on non-Java clients and custom protocols is covered in Apache HBase External APIs

67. Client Request Filters

Get and Scan instances can be optionally configured with filters which are applied on the RegionServer.

Filters can be confusing because there are many different types, and it is best to approach them by understanding the groups of Filter functionality.

67.1. Structural

Structural Filters contain other Filters.

67.1.1. FilterList

FilterList represents a list of Filters with a relationship of FilterList.Operator.MUST_PASS_ALL or FilterList.Operator.MUST_PASS_ONE between the Filters. The following example shows an 'or' between two Filters (checking for either 'my value' or 'my other value' on the same attribute).

FilterList list = new FilterList(FilterList.Operator.MUST_PASS_ONE);
SingleColumnValueFilter filter1 = new SingleColumnValueFilter(
  Bytes.toBytes("my value")
SingleColumnValueFilter filter2 = new SingleColumnValueFilter(
  Bytes.toBytes("my other value")

67.2. Column Value

67.2.1. SingleColumnValueFilter

A SingleColumnValueFilter (see: http://hbase.apache.org/apidocs/org/apache/hadoop/hbase/filter/SingleColumnValueFilter.html) can be used to test column values for equivalence (CompareOp.EQUAL), inequality (CompareOp.NOT_EQUAL), or ranges (e.g., CompareOp.GREATER). The following is an example of testing equivalence of a column to a String value "my value"…​

SingleColumnValueFilter filter = new SingleColumnValueFilter(
  Bytes.toBytes("my value")

67.3. Column Value Comparators

There are several Comparator classes in the Filter package that deserve special mention. These Comparators are used in concert with other Filters, such as SingleColumnValueFilter.

67.3.1. RegexStringComparator

RegexStringComparator supports regular expressions for value comparisons.

RegexStringComparator comp = new RegexStringComparator("my.");   // any value that starts with 'my'
SingleColumnValueFilter filter = new SingleColumnValueFilter(

See the Oracle JavaDoc for supported RegEx patterns in Java.

67.3.2. SubstringComparator

SubstringComparator can be used to determine if a given substring exists in a value. The comparison is case-insensitive.

SubstringComparator comp = new SubstringComparator("y val");   // looking for 'my value'
SingleColumnValueFilter filter = new SingleColumnValueFilter(

67.3.3. BinaryPrefixComparator

67.3.4. BinaryComparator

67.4. KeyValue Metadata

As HBase stores data internally as KeyValue pairs, KeyValue Metadata Filters evaluate the existence of keys (i.e., ColumnFamily:Column qualifiers) for a row, as opposed to values the previous section.

67.4.1. FamilyFilter

FamilyFilter can be used to filter on the ColumnFamily. It is generally a better idea to select ColumnFamilies in the Scan than to do it with a Filter.

67.4.2. QualifierFilter

QualifierFilter can be used to filter based on Column (aka Qualifier) name.

67.4.3. ColumnPrefixFilter

ColumnPrefixFilter can be used to filter based on the lead portion of Column (aka Qualifier) names.

A ColumnPrefixFilter seeks ahead to the first column matching the prefix in each row and for each involved column family. It can be used to efficiently get a subset of the columns in very wide rows.

Note: The same column qualifier can be used in different column families. This filter returns all matching columns.

Example: Find all columns in a row and family that start with "abc"

HTableInterface t = ...;
byte[] row = ...;
byte[] family = ...;
byte[] prefix = Bytes.toBytes("abc");
Scan scan = new Scan(row, row); // (optional) limit to one row
scan.addFamily(family); // (optional) limit to one family
Filter f = new ColumnPrefixFilter(prefix);
scan.setBatch(10); // set this if there could be many columns returned
ResultScanner rs = t.getScanner(scan);
for (Result r = rs.next(); r != null; r = rs.next()) {
  for (KeyValue kv : r.raw()) {
    // each kv represents a column

67.4.4. MultipleColumnPrefixFilter

MultipleColumnPrefixFilter behaves like ColumnPrefixFilter but allows specifying multiple prefixes.

Like ColumnPrefixFilter, MultipleColumnPrefixFilter efficiently seeks ahead to the first column matching the lowest prefix and also seeks past ranges of columns between prefixes. It can be used to efficiently get discontinuous sets of columns from very wide rows.

Example: Find all columns in a row and family that start with "abc" or "xyz"

HTableInterface t = ...;
byte[] row = ...;
byte[] family = ...;
byte[][] prefixes = new byte[][] {Bytes.toBytes("abc"), Bytes.toBytes("xyz")};
Scan scan = new Scan(row, row); // (optional) limit to one row
scan.addFamily(family); // (optional) limit to one family
Filter f = new MultipleColumnPrefixFilter(prefixes);
scan.setBatch(10); // set this if there could be many columns returned
ResultScanner rs = t.getScanner(scan);
for (Result r = rs.next(); r != null; r = rs.next()) {
  for (KeyValue kv : r.raw()) {
    // each kv represents a column

67.4.5. ColumnRangeFilter

A ColumnRangeFilter allows efficient intra row scanning.

A ColumnRangeFilter can seek ahead to the first matching column for each involved column family. It can be used to efficiently get a 'slice' of the columns of a very wide row. i.e. you have a million columns in a row but you only want to look at columns bbbb-bbdd.

Note: The same column qualifier can be used in different column families. This filter returns all matching columns.

Example: Find all columns in a row and family between "bbbb" (inclusive) and "bbdd" (inclusive)

HTableInterface t = ...;
byte[] row = ...;
byte[] family = ...;
byte[] startColumn = Bytes.toBytes("bbbb");
byte[] endColumn = Bytes.toBytes("bbdd");
Scan scan = new Scan(row, row); // (optional) limit to one row
scan.addFamily(family); // (optional) limit to one family
Filter f = new ColumnRangeFilter(startColumn, true, endColumn, true);
scan.setBatch(10); // set this if there could be many columns returned
ResultScanner rs = t.getScanner(scan);
for (Result r = rs.next(); r != null; r = rs.next()) {
  for (KeyValue kv : r.raw()) {
    // each kv represents a column

Note: Introduced in HBase 0.92

67.5. RowKey

67.5.1. RowFilter

It is generally a better idea to use the startRow/stopRow methods on Scan for row selection, however RowFilter can also be used.

67.6. Utility

67.6.1. FirstKeyOnlyFilter

This is primarily used for rowcount jobs. See FirstKeyOnlyFilter.

68. Master

HMaster is the implementation of the Master Server. The Master server is responsible for monitoring all RegionServer instances in the cluster, and is the interface for all metadata changes. In a distributed cluster, the Master typically runs on the NameNode. J Mohamed Zahoor goes into some more detail on the Master Architecture in this blog posting, HBase HMaster Architecture .

68.1. Startup Behavior

If run in a multi-Master environment, all Masters compete to run the cluster. If the active Master loses its lease in ZooKeeper (or the Master shuts down), then the remaining Masters jostle to take over the Master role.

68.2. Runtime Impact

A common dist-list question involves what happens to an HBase cluster when the Master goes down. Because the HBase client talks directly to the RegionServers, the cluster can still function in a "steady state". Additionally, per Catalog Tables, hbase:meta exists as an HBase table and is not resident in the Master. However, the Master controls critical functions such as RegionServer failover and completing region splits. So while the cluster can still run for a short time without the Master, the Master should be restarted as soon as possible.

68.3. Interface

The methods exposed by HMasterInterface are primarily metadata-oriented methods:

  • Table (createTable, modifyTable, removeTable, enable, disable)

  • ColumnFamily (addColumn, modifyColumn, removeColumn)

  • Region (move, assign, unassign) For example, when the Admin method disableTable is invoked, it is serviced by the Master server.

68.4. Processes

The Master runs several background threads:

68.4.1. LoadBalancer

Periodically, and when there are no regions in transition, a load balancer will run and move regions around to balance the cluster’s load. See Balancer for configuring this property.

See Region-RegionServer Assignment for more information on region assignment.

68.4.2. CatalogJanitor

Periodically checks and cleans up the hbase:meta table. See <arch.catalog.meta>> for more information on the meta table.

69. RegionServer

HRegionServer is the RegionServer implementation. It is responsible for serving and managing regions. In a distributed cluster, a RegionServer runs on a DataNode.

69.1. Interface

The methods exposed by HRegionRegionInterface contain both data-oriented and region-maintenance methods:

  • Data (get, put, delete, next, etc.)

  • Region (splitRegion, compactRegion, etc.) For example, when the Admin method majorCompact is invoked on a table, the client is actually iterating through all regions for the specified table and requesting a major compaction directly to each region.

69.2. Processes

The RegionServer runs a variety of background threads:

69.2.1. CompactSplitThread

Checks for splits and handle minor compactions.

69.2.2. MajorCompactionChecker

Checks for major compactions.

69.2.3. MemStoreFlusher

Periodically flushes in-memory writes in the MemStore to StoreFiles.

69.2.4. LogRoller

Periodically checks the RegionServer’s WAL.

69.3. Coprocessors

Coprocessors were added in 0.92. There is a thorough Blog Overview of CoProcessors posted. Documentation will eventually move to this reference guide, but the blog is the most current information available at this time.

69.4. Block Cache

HBase provides two different BlockCache implementations: the default on-heap LruBlockCache and the BucketCache, which is (usually) off-heap. This section discusses benefits and drawbacks of each implementation, how to choose the appropriate option, and configuration options for each.

Block Cache Reporting: UI

See the RegionServer UI for detail on caching deploy. Since HBase 0.98.4, the Block Cache detail has been significantly extended showing configurations, sizings, current usage, time-in-the-cache, and even detail on block counts and types.

69.4.1. Cache Choices

LruBlockCache is the original implementation, and is entirely within the Java heap. BucketCache is mainly intended for keeping block cache data off-heap, although BucketCache can also keep data on-heap and serve from a file-backed cache.

BucketCache is production ready as of HBase 0.98.6

To run with BucketCache, you need HBASE-11678. This was included in 0.98.6.

Fetching will always be slower when fetching from BucketCache, as compared to the native on-heap LruBlockCache. However, latencies tend to be less erratic across time, because there is less garbage collection when you use BucketCache since it is managing BlockCache allocations, not the GC. If the BucketCache is deployed in off-heap mode, this memory is not managed by the GC at all. This is why you’d use BucketCache, so your latencies are less erratic and to mitigate GCs and heap fragmentation. See Nick Dimiduk’s BlockCache 101 for comparisons running on-heap vs off-heap tests. Also see Comparing BlockCache Deploys which finds that if your dataset fits inside your LruBlockCache deploy, use it otherwise if you are experiencing cache churn (or you want your cache to exist beyond the vagaries of java GC), use BucketCache.

When you enable BucketCache, you are enabling a two tier caching system, an L1 cache which is implemented by an instance of LruBlockCache and an off-heap L2 cache which is implemented by BucketCache. Management of these two tiers and the policy that dictates how blocks move between them is done by CombinedBlockCache. It keeps all DATA blocks in the L2 BucketCache and meta blocks — INDEX and BLOOM blocks — on-heap in the L1 LruBlockCache. See Off-heap Block Cache for more detail on going off-heap.

69.4.2. General Cache Configurations

Apart from the cache implementation itself, you can set some general configuration options to control how the cache performs. See http://hbase.apache.org/devapidocs/org/apache/hadoop/hbase/io/hfile/CacheConfig.html. After setting any of these options, restart or rolling restart your cluster for the configuration to take effect. Check logs for errors or unexpected behavior.

See also Prefetch Option for Blockcache, which discusses a new option introduced in HBASE-9857.

69.4.3. LruBlockCache Design

The LruBlockCache is an LRU cache that contains three levels of block priority to allow for scan-resistance and in-memory ColumnFamilies:

  • Single access priority: The first time a block is loaded from HDFS it normally has this priority and it will be part of the first group to be considered during evictions. The advantage is that scanned blocks are more likely to get evicted than blocks that are getting more usage.

  • Multi access priority: If a block in the previous priority group is accessed again, it upgrades to this priority. It is thus part of the second group considered during evictions.

  • In-memory access priority: If the block’s family was configured to be "in-memory", it will be part of this priority disregarding the number of times it was accessed. Catalog tables are configured like this. This group is the last one considered during evictions.

    To mark a column family as in-memory, call


if creating a table from java, or set IN_MEMORY ⇒ true when creating or altering a table in the shell: e.g.

hbase(main):003:0> create  't', {NAME => 'f', IN_MEMORY => 'true'}

For more information, see the LruBlockCache source

69.4.4. LruBlockCache Usage

Block caching is enabled by default for all the user tables which means that any read operation will load the LRU cache. This might be good for a large number of use cases, but further tunings are usually required in order to achieve better performance. An important concept is the working set size, or WSS, which is: "the amount of memory needed to compute the answer to a problem". For a website, this would be the data that’s needed to answer the queries over a short amount of time.

The way to calculate how much memory is available in HBase for caching is:

number of region servers * heap size * hfile.block.cache.size * 0.99

The default value for the block cache is 0.25 which represents 25% of the available heap. The last value (99%) is the default acceptable loading factor in the LRU cache after which eviction is started. The reason it is included in this equation is that it would be unrealistic to say that it is possible to use 100% of the available memory since this would make the process blocking from the point where it loads new blocks. Here are some examples:

  • One region server with the heap size set to 1 GB and the default block cache size will have 253 MB of block cache available.

  • 20 region servers with the heap size set to 8 GB and a default block cache size will have 39.6 of block cache.

  • 100 region servers with the heap size set to 24 GB and a block cache size of 0.5 will have about 1.16 TB of block cache.

Your data is not the only resident of the block cache. Here are others that you may have to take into account:

Catalog Tables

The -ROOT- (prior to HBase 0.96, see arch.catalog.root) and hbase:meta tables are forced into the block cache and have the in-memory priority which means that they are harder to evict. The former never uses more than a few hundred bytes while the latter can occupy a few MBs (depending on the number of regions).

HFiles Indexes

An HFile is the file format that HBase uses to store data in HDFS. It contains a multi-layered index which allows HBase to seek to the data without having to read the whole file. The size of those indexes is a factor of the block size (64KB by default), the size of your keys and the amount of data you are storing. For big data sets it’s not unusual to see numbers around 1GB per region server, although not all of it will be in cache because the LRU will evict indexes that aren’t used.


The values that are stored are only half the picture, since each value is stored along with its keys (row key, family qualifier, and timestamp). See Try to minimize row and column sizes.

Bloom Filters

Just like the HFile indexes, those data structures (when enabled) are stored in the LRU.

Currently the recommended way to measure HFile indexes and bloom filters sizes is to look at the region server web UI and checkout the relevant metrics. For keys, sampling can be done by using the HFile command line tool and look for the average key size metric. Since HBase 0.98.3, you can view details on BlockCache stats and metrics in a special Block Cache section in the UI.

It’s generally bad to use block caching when the WSS doesn’t fit in memory. This is the case when you have for example 40GB available across all your region servers' block caches but you need to process 1TB of data. One of the reasons is that the churn generated by the evictions will trigger more garbage collections unnecessarily. Here are two use cases:

  • Fully random reading pattern: This is a case where you almost never access the same row twice within a short amount of time such that the chance of hitting a cached block is close to 0. Setting block caching on such a table is a waste of memory and CPU cycles, more so that it will generate more garbage to pick up by the JVM. For more information on monitoring GC, see JVM Garbage Collection Logs.

  • Mapping a table: In a typical MapReduce job that takes a table in input, every row will be read only once so there’s no need to put them into the block cache. The Scan object has the option of turning this off via the setCaching method (set it to false). You can still keep block caching turned on on this table if you need fast random read access. An example would be counting the number of rows in a table that serves live traffic, caching every block of that table would create massive churn and would surely evict data that’s currently in use.

Caching META blocks only (DATA blocks in fscache)

An interesting setup is one where we cache META blocks only and we read DATA blocks in on each access. If the DATA blocks fit inside fscache, this alternative may make sense when access is completely random across a very large dataset. To enable this setup, alter your table and for each column family set BLOCKCACHE ⇒ 'false'. You are 'disabling' the BlockCache for this column family only. You can never disable the caching of META blocks. Since HBASE-4683 Always cache index and bloom blocks, we will cache META blocks even if the BlockCache is disabled.

69.4.5. Off-heap Block Cache

How to Enable BucketCache

The usual deploy of BucketCache is via a managing class that sets up two caching tiers: an L1 on-heap cache implemented by LruBlockCache and a second L2 cache implemented with BucketCache. The managing class is CombinedBlockCache by default. The previous link describes the caching 'policy' implemented by CombinedBlockCache. In short, it works by keeping meta blocks — INDEX and BLOOM in the L1, on-heap LruBlockCache tier — and DATA blocks are kept in the L2, BucketCache tier. It is possible to amend this behavior in HBase since version 1.0 and ask that a column family have both its meta and DATA blocks hosted on-heap in the L1 tier by setting cacheDataInL1 via (HColumnDescriptor.setCacheDataInL1(true) or in the shell, creating or amending column families setting CACHE_DATA_IN_L1 to true: e.g.

hbase(main):003:0> create 't', {NAME => 't', CONFIGURATION => {CACHE_DATA_IN_L1 => 'true'}}

The BucketCache Block Cache can be deployed on-heap, off-heap, or file based. You set which via the hbase.bucketcache.ioengine setting. Setting it to heap will have BucketCache deployed inside the allocated Java heap. Setting it to offheap will have BucketCache make its allocations off-heap, and an ioengine setting of file:PATH_TO_FILE will direct BucketCache to use a file caching (Useful in particular if you have some fast I/O attached to the box such as SSDs).

It is possible to deploy an L1+L2 setup where we bypass the CombinedBlockCache policy and have BucketCache working as a strict L2 cache to the L1 LruBlockCache. For such a setup, set CacheConfig.BUCKET_CACHE_COMBINED_KEY to false. In this mode, on eviction from L1, blocks go to L2. When a block is cached, it is cached first in L1. When we go to look for a cached block, we look first in L1 and if none found, then search L2. Let us call this deploy format, Raw L1+L2.

Other BucketCache configs include: specifying a location to persist cache to across restarts, how many threads to use writing the cache, etc. See the CacheConfig.html class for configuration options and descriptions.

BucketCache Example Configuration

This sample provides a configuration for a 4 GB off-heap BucketCache with a 1 GB on-heap cache.

Configuration is performed on the RegionServer.

Setting hbase.bucketcache.ioengine and hbase.bucketcache.size > 0 enables CombinedBlockCache. Let us presume that the RegionServer has been set to run with a 5G heap: i.e. HBASE_HEAPSIZE=5g.

  1. First, edit the RegionServer’s hbase-env.sh and set HBASE_OFFHEAPSIZE to a value greater than the off-heap size wanted, in this case, 4 GB (expressed as 4G). Let’s set it to 5G. That’ll be 4G for our off-heap cache and 1G for any other uses of off-heap memory (there are other users of off-heap memory other than BlockCache; e.g. DFSClient in RegionServer can make use of off-heap memory). See Direct Memory Usage In HBase.

  2. Next, add the following configuration to the RegionServer’s hbase-site.xml.

  3. Restart or rolling restart your cluster, and check the logs for any issues.

In the above, we set the BucketCache to be 4G. We configured the on-heap LruBlockCache have 20% (0.2) of the RegionServer’s heap size (0.2 * 5G = 1G). In other words, you configure the L1 LruBlockCache as you would normally (as if there were no L2 cache present).

HBASE-10641 introduced the ability to configure multiple sizes for the buckets of the BucketCache, in HBase 0.98 and newer. To configurable multiple bucket sizes, configure the new property hbase.bucketcache.bucket.sizes to a comma-separated list of block sizes, ordered from smallest to largest, with no spaces. The goal is to optimize the bucket sizes based on your data access patterns. The following example configures buckets of size 4096 and 8192.

Direct Memory Usage In HBase

The default maximum direct memory varies by JVM. Traditionally it is 64M or some relation to allocated heap size (-Xmx) or no limit at all (JDK7 apparently). HBase servers use direct memory, in particular short-circuit reading, the hosted DFSClient will allocate direct memory buffers. If you do off-heap block caching, you’ll be making use of direct memory. Starting your JVM, make sure the -XX:MaxDirectMemorySize setting in conf/hbase-env.sh is set to some value that is higher than what you have allocated to your off-heap BlockCache (hbase.bucketcache.size). It should be larger than your off-heap block cache and then some for DFSClient usage (How much the DFSClient uses is not easy to quantify; it is the number of open HFiles * hbase.dfs.client.read.shortcircuit.buffer.size where hbase.dfs.client.read.shortcircuit.buffer.size is set to 128k in HBase — see hbase-default.xml default configurations). Direct memory, which is part of the Java process heap, is separate from the object heap allocated by -Xmx. The value allocated by MaxDirectMemorySize must not exceed physical RAM, and is likely to be less than the total available RAM due to other memory requirements and system constraints.

You can see how much memory — on-heap and off-heap/direct — a RegionServer is configured to use and how much it is using at any one time by looking at the Server Metrics: Memory tab in the UI. It can also be gotten via JMX. In particular the direct memory currently used by the server can be found on the java.nio.type=BufferPool,name=direct bean. Terracotta has a good write up on using off-heap memory in Java. It is for their product BigMemory but a lot of the issues noted apply in general to any attempt at going off-heap. Check it out.


This is a pre-HBase 1.0 configuration removed because it was confusing. It was a float that you would set to some value between 0.0 and 1.0. Its default was 0.9. If the deploy was using CombinedBlockCache, then the LruBlockCache L1 size was calculated to be (1 - hbase.bucketcache.percentage.in.combinedcache) * size-of-bucketcache and the BucketCache size was hbase.bucketcache.percentage.in.combinedcache * size-of-bucket-cache. where size-of-bucket-cache itself is EITHER the value of the configuration hbase.bucketcache.size IF it was specified as Megabytes OR hbase.bucketcache.size * -XX:MaxDirectMemorySize if hbase.bucketcache.size is between 0 and 1.0.

In 1.0, it should be more straight-forward. L1 LruBlockCache size is set as a fraction of java heap using hfile.block.cache.size setting (not the best name) and L2 is set as above either in absolute Megabytes or as a fraction of allocated maximum direct memory.

69.4.6. Compressed BlockCache

HBASE-11331 introduced lazy BlockCache decompression, more simply referred to as compressed BlockCache. When compressed BlockCache is enabled data and encoded data blocks are cached in the BlockCache in their on-disk format, rather than being decompressed and decrypted before caching.

For a RegionServer hosting more data than can fit into cache, enabling this feature with SNAPPY compression has been shown to result in 50% increase in throughput and 30% improvement in mean latency while, increasing garbage collection by 80% and increasing overall CPU load by 2%. See HBASE-11331 for more details about how performance was measured and achieved. For a RegionServer hosting data that can comfortably fit into cache, or if your workload is sensitive to extra CPU or garbage-collection load, you may receive less benefit.

The compressed BlockCache is disabled by default. To enable it, set hbase.block.data.cachecompressed to true in hbase-site.xml on all RegionServers.

69.5. RegionServer Splitting Implementation

As write requests are handled by the region server, they accumulate in an in-memory storage system called the memstore. Once the memstore fills, its content are written to disk as additional store files. This event is called a memstore flush. As store files accumulate, the RegionServer will compact them into fewer, larger files. After each flush or compaction finishes, the amount of data stored in the region has changed. The RegionServer consults the region split policy to determine if the region has grown too large or should be split for another policy-specific reason. A region split request is enqueued if the policy recommends it.

Logically, the process of splitting a region is simple. We find a suitable point in the keyspace of the region where we should divide the region in half, then split the region’s data into two new regions at that point. The details of the process however are not simple. When a split happens, the newly created daughter regions do not rewrite all the data into new files immediately. Instead, they create small files similar to symbolic link files, named Reference files, which point to either the top or bottom part of the parent store file according to the split point. The reference file is used just like a regular data file, but only half of the records are considered. The region can only be split if there are no more references to the immutable data files of the parent region. Those reference files are cleaned gradually by compactions, so that the region will stop referring to its parents files, and can be split further.

Although splitting the region is a local decision made by the RegionServer, the split process itself must coordinate with many actors. The RegionServer notifies the Master before and after the split, updates the .META. table so that clients can discover the new daughter regions, and rearranges the directory structure and data files in HDFS. Splitting is a multi-task process. To enable rollback in case of an error, the RegionServer keeps an in-memory journal about the execution state. The steps taken by the RegionServer to execute the split are illustrated in RegionServer Split Process. Each step is labeled with its step number. Actions from RegionServers or Master are shown in red, while actions from the clients are show in green.

Region Split Process
Figure 1. RegionServer Split Process
  1. The RegionServer decides locally to split the region, and prepares the split. THE SPLIT TRANSACTION IS STARTED. As a first step, the RegionServer acquires a shared read lock on the table to prevent schema modifications during the splitting process. Then it creates a znode in zookeeper under /hbase/region-in-transition/region-name, and sets the znode’s state to SPLITTING.

  2. The Master learns about this znode, since it has a watcher for the parent region-in-transition znode.

  3. The RegionServer creates a sub-directory named .splits under the parent’s region directory in HDFS.

  4. The RegionServer closes the parent region and marks the region as offline in its local data structures. THE SPLITTING REGION IS NOW OFFLINE. At this point, client requests coming to the parent region will throw NotServingRegionException. The client will retry with some backoff. The closing region is flushed.

  5. The RegionServer creates region directories under the .splits directory, for daughter regions A and B, and creates necessary data structures. Then it splits the store files, in the sense that it creates two Reference files per store file in the parent region. Those reference files will point to the parent region’s files.

  6. The RegionServer creates the actual region directory in HDFS, and moves the reference files for each daughter.

  7. The RegionServer sends a Put request to the .META. table, to set the parent as offline in the .META. table and add information about daughter regions. At this point, there won’t be individual entries in .META. for the daughters. Clients will see that the parent region is split if they scan .META., but won’t know about the daughters until they appear in .META.. Also, if this Put to .META. succeeds, the parent will be effectively split. If the RegionServer fails before this RPC succeeds, Master and the next Region Server opening the region will clean dirty state about the region split. After the .META. update, though, the region split will be rolled-forward by Master.

  8. The RegionServer opens daughters A and B in parallel.

  9. The RegionServer adds the daughters A and B to .META., together with information that it hosts the regions. THE SPLIT REGIONS (DAUGHTERS WITH REFERENCES TO PARENT) ARE NOW ONLINE. After this point, clients can discover the new regions and issue requests to them. Clients cache the .META. entries locally, but when they make requests to the RegionServer or .META., their caches will be invalidated, and they will learn about the new regions from .META..

  10. The RegionServer updates znode /hbase/region-in-transition/region-name in ZooKeeper to state SPLIT, so that the master can learn about it. The balancer can freely re-assign the daughter regions to other region servers if necessary. THE SPLIT TRANSACTION IS NOW FINISHED.

  11. After the split, .META. and HDFS will still contain references to the parent region. Those references will be removed when compactions in daughter regions rewrite the data files. Garbage collection tasks in the master periodically check whether the daughter regions still refer to the parent region’s files. If not, the parent region will be removed.

69.6. Write Ahead Log (WAL)

69.6.1. Purpose

The Write Ahead Log (WAL) records all changes to data in HBase, to file-based storage. Under normal operations, the WAL is not needed because data changes move from the MemStore to StoreFiles. However, if a RegionServer crashes or becomes unavailable before the MemStore is flushed, the WAL ensures that the changes to the data can be replayed. If writing to the WAL fails, the entire operation to modify the data fails.

HBase uses an implementation of the WAL interface. Usually, there is only one instance of a WAL per RegionServer. The RegionServer records Puts and Deletes to it, before recording them to the MemStore for the affected Store.

The HLog

Prior to 2.0, the interface for WALs in HBase was named HLog. In 0.94, HLog was the name of the implementation of the WAL. You will likely find references to the HLog in documentation tailored to these older versions.

The WAL resides in HDFS in the /hbase/WALs/ directory (prior to HBase 0.94, they were stored in /hbase/.logs/), with subdirectories per region.

For more general information about the concept of write ahead logs, see the Wikipedia Write-Ahead Log article.

69.6.2. MultiWAL

With a single WAL per RegionServer, the RegionServer must write to the WAL serially, because HDFS files must be sequential. This causes the WAL to be a performance bottleneck.

HBase 1.0 introduces support MultiWal in HBASE-5699. MultiWAL allows a RegionServer to write multiple WAL streams in parallel, by using multiple pipelines in the underlying HDFS instance, which increases total throughput during writes. This parallelization is done by partitioning incoming edits by their Region. Thus, the current implementation will not help with increasing the throughput to a single Region.

RegionServers using the original WAL implementation and those using the MultiWAL implementation can each handle recovery of either set of WALs, so a zero-downtime configuration update is possible through a rolling restart.

Configure MultiWAL

To configure MultiWAL for a RegionServer, set the value of the property hbase.wal.provider to multiwal by pasting in the following XML:


Restart the RegionServer for the changes to take effect.

To disable MultiWAL for a RegionServer, unset the property and restart the RegionServer.

69.6.3. WAL Flushing

TODO (describe).

69.6.4. WAL Splitting

A RegionServer serves many regions. All of the regions in a region server share the same active WAL file. Each edit in the WAL file includes information about which region it belongs to. When a region is opened, the edits in the WAL file which belong to that region need to be replayed. Therefore, edits in the WAL file must be grouped by region so that particular sets can be replayed to regenerate the data in a particular region. The process of grouping the WAL edits by region is called log splitting. It is a critical process for recovering data if a region server fails.

Log splitting is done by the HMaster during cluster start-up or by the ServerShutdownHandler as a region server shuts down. So that consistency is guaranteed, affected regions are unavailable until data is restored. All WAL edits need to be recovered and replayed before a given region can become available again. As a result, regions affected by log splitting are unavailable until the process completes.

Procedure: Log Splitting, Step by Step
  1. The /hbase/WALs/<host>,<port>,<startcode> directory is renamed.

    Renaming the directory is important because a RegionServer may still be up and accepting requests even if the HMaster thinks it is down. If the RegionServer does not respond immediately and does not heartbeat its ZooKeeper session, the HMaster may interpret this as a RegionServer failure. Renaming the logs directory ensures that existing, valid WAL files which are still in use by an active but busy RegionServer are not written to by accident.

    The new directory is named according to the following pattern:


    An example of such a renamed directory might look like the following:

  2. Each log file is split, one at a time.

    The log splitter reads the log file one edit entry at a time and puts each edit entry into the buffer corresponding to the edit’s region. At the same time, the splitter starts several writer threads. Writer threads pick up a corresponding buffer and write the edit entries in the buffer to a temporary recovered edit file. The temporary edit file is stored to disk with the following naming pattern:


    This file is used to store all the edits in the WAL log for this region. After log splitting completes, the .temp file is renamed to the sequence ID of the first log written to the file.

    To determine whether all edits have been written, the sequence ID is compared to the sequence of the last edit that was written to the HFile. If the sequence of the last edit is greater than or equal to the sequence ID included in the file name, it is clear that all writes from the edit file have been completed.

  3. After log splitting is complete, each affected region is assigned to a RegionServer.

    When the region is opened, the recovered.edits folder is checked for recovered edits files. If any such files are present, they are replayed by reading the edits and saving them to the MemStore. After all edit files are replayed, the contents of the MemStore are written to disk (HFile) and the edit files are deleted.

Handling of Errors During Log Splitting

If you set the hbase.hlog.split.skip.errors option to true, errors are treated as follows:

  • Any error encountered during splitting will be logged.

  • The problematic WAL log will be moved into the .corrupt directory under the hbase rootdir,

  • Processing of the WAL will continue

If the hbase.hlog.split.skip.errors option is set to false, the default, the exception will be propagated and the split will be logged as failed. See HBASE-2958 When hbase.hlog.split.skip.errors is set to false, we fail the split but that’s it. We need to do more than just fail split if this flag is set.

How EOFExceptions are treated when splitting a crashed RegionServer’s WALs

If an EOFException occurs while splitting logs, the split proceeds even when hbase.hlog.split.skip.errors is set to false. An EOFException while reading the last log in the set of files to split is likely, because the RegionServer was likely in the process of writing a record at the time of a crash. For background, see HBASE-2643 Figure how to deal with eof splitting logs

Performance Improvements during Log Splitting

WAL log splitting and recovery can be resource intensive and take a long time, depending on the number of RegionServers involved in the crash and the size of the regions. [distributed.log.splitting] was developed to improve performance during log splitting.

Enabling or Disabling Distributed Log Splitting

Distributed log processing is enabled by default since HBase 0.92. The setting is controlled by the hbase.master.distributed.log.splitting property, which can be set to true or false, but defaults to true.

Distributed Log Splitting, Step by Step

After configuring distributed log splitting, the HMaster controls the process. The HMaster enrolls each RegionServer in the log splitting process, and the actual work of splitting the logs is done by the RegionServers. The general process for log splitting, as described in Distributed Log Splitting, Step by Step still applies here.

  1. If distributed log processing is enabled, the HMaster creates a split log manager instance when the cluster is started.

    1. The split log manager manages all log files which need to be scanned and split.

    2. The split log manager places all the logs into the ZooKeeper splitlog node (/hbase/splitlog) as tasks.

    3. You can view the contents of the splitlog by issuing the following zkCli command. Example output is shown.

      ls /hbase/splitlog

      The output contains some non-ASCII characters. When decoded, it looks much more simple:


      The listing represents WAL file names to be scanned and split, which is a list of log splitting tasks.

  2. The split log manager monitors the log-splitting tasks and workers.

    The split log manager is responsible for the following ongoing tasks:

    • Once the split log manager publishes all the tasks to the splitlog znode, it monitors these task nodes and waits for them to be processed.

    • Checks to see if there are any dead split log workers queued up. If it finds tasks claimed by unresponsive workers, it will resubmit those tasks. If the resubmit fails due to some ZooKeeper exception, the dead worker is queued up again for retry.

    • Checks to see if there are any unassigned tasks. If it finds any, it create an ephemeral rescan node so that each split log worker is notified to re-scan unassigned tasks via the nodeChildrenChanged ZooKeeper event.

    • Checks for tasks which are assigned but expired. If any are found, they are moved back to TASK_UNASSIGNED state again so that they can be retried. It is possible that these tasks are assigned to slow workers, or they may already be finished. This is not a problem, because log splitting tasks have the property of idempotence. In other words, the same log splitting task can be processed many times without causing any problem.

    • The split log manager watches the HBase split log znodes constantly. If any split log task node data is changed, the split log manager retrieves the node data. The node data contains the current state of the task. You can use the zkCli get command to retrieve the current state of a task. In the example output below, the first line of the output shows that the task is currently unassigned.

      get /hbase/splitlog/hdfs%3A%2F%2Fhost2.sample.com%3A56020%2Fhbase%2F.logs%2Fhost6.sample.com%2C57020%2C1340474893287-splitting%2Fhost6.sample.com%253A57020.1340474893945
      unassigned host2.sample.com:57000
      cZxid = 0×7115
      ctime = Sat Jun 23 11:13:40 PDT 2012

      Based on the state of the task whose data is changed, the split log manager does one of the following:

    • Resubmit the task if it is unassigned

    • Heartbeat the task if it is assigned

    • Resubmit or fail the task if it is resigned (see Reasons a Task Will Fail)

    • Resubmit or fail the task if it is completed with errors (see Reasons a Task Will Fail)

    • Resubmit or fail the task if it could not complete due to errors (see Reasons a Task Will Fail)

    • Delete the task if it is successfully completed or failed

      Reasons a Task Will Fail
      • The task has been deleted.

      • The node no longer exists.

      • The log status manager failed to move the state of the task to TASK_UNASSIGNED.

      • The number of resubmits is over the resubmit threshold.

  3. Each RegionServer’s split log worker performs the log-splitting tasks.

    Each RegionServer runs a daemon thread called the split log worker, which does the work to split the logs. The daemon thread starts when the RegionServer starts, and registers itself to watch HBase znodes. If any splitlog znode children change, it notifies a sleeping worker thread to wake up and grab more tasks. If a worker’s current task’s node data is changed, the worker checks to see if the task has been taken by another worker. If so, the worker thread stops work on the current task.

    The worker monitors the splitlog znode constantly. When a new task appears, the split log worker retrieves the task paths and checks each one until it finds an unclaimed task, which it attempts to claim. If the claim was successful, it attempts to perform the task and updates the task’s state property based on the splitting outcome. At this point, the split log worker scans for another unclaimed task.

    How the Split Log Worker Approaches a Task
    • It queries the task state and only takes action if the task is in `TASK_UNASSIGNED `state.

    • If the task is in TASK_UNASSIGNED state, the worker attempts to set the state to TASK_OWNED by itself. If it fails to set the state, another worker will try to grab it. The split log manager will also ask all workers to rescan later if the task remains unassigned.

    • If the worker succeeds in taking ownership of the task, it tries to get the task state again to make sure it really gets it asynchronously. In the meantime, it starts a split task executor to do the actual work:

      • Get the HBase root folder, create a temp folder under the root, and split the log file to the temp folder.

      • If the split was successful, the task executor sets the task to state TASK_DONE.

      • If the worker catches an unexpected IOException, the task is set to state TASK_ERR.

      • If the worker is shutting down, set the task to state TASK_RESIGNED.

      • If the task is taken by another worker, just log it.

  4. The split log manager monitors for uncompleted tasks.

    The split log manager returns when all tasks are completed successfully. If all tasks are completed with some failures, the split log manager throws an exception so that the log splitting can be retried. Due to an asynchronous implementation, in very rare cases, the split log manager loses track of some completed tasks. For that reason, it periodically checks for remaining uncompleted task in its task map or ZooKeeper. If none are found, it throws an exception so that the log splitting can be retried right away instead of hanging there waiting for something that won’t happen.

Distributed Log Replay

After a RegionServer fails, its failed regions are assigned to another RegionServer, which are marked as "recovering" in ZooKeeper. A split log worker directly replays edits from the WAL of the failed RegionServer to the regions at its new location. When a region is in "recovering" state, it can accept writes but no reads (including Append and Increment), region splits or merges.

Distributed Log Replay extends the [distributed.log.splitting] framework. It works by directly replaying WAL edits to another RegionServer instead of creating recovered.edits files. It provides the following advantages over distributed log splitting alone:

  • It eliminates the overhead of writing and reading a large number of recovered.edits files. It is not unusual for thousands of recovered.edits files to be created and written concurrently during a RegionServer recovery. Many small random writes can degrade overall system performance.

  • It allows writes even when a region is in recovering state. It only takes seconds for a recovering region to accept writes again.

Enabling Distributed Log Replay

To enable distributed log replay, set hbase.master.distributed.log.replay to true. This will be the default for HBase 0.99 (HBASE-10888).

You must also enable HFile version 3 (which is the default HFile format starting in HBase 0.99. See HBASE-10855). Distributed log replay is unsafe for rolling upgrades.

69.6.5. Disabling the WAL

It is possible to disable the WAL, to improve performance in certain specific situations. However, disabling the WAL puts your data at risk. The only situation where this is recommended is during a bulk load. This is because, in the event of a problem, the bulk load can be re-run with no risk of data loss.

The WAL is disabled by calling the HBase client field Mutation.writeToWAL(false). Use the Mutation.setDurability(Durability.SKIP_WAL) and Mutation.getDurability() methods to set and get the field’s value. There is no way to disable the WAL for only a specific table.

If you disable the WAL for anything other than bulk loads, your data is at risk.

70. Regions

Regions are the basic element of availability and distribution for tables, and are comprised of a Store per Column Family. The hierarchy of objects is as follows:

Table                    (HBase table)
    Region               (Regions for the table)
        Store            (Store per ColumnFamily for each Region for the table)
            MemStore     (MemStore for each Store for each Region for the table)
            StoreFile    (StoreFiles for each Store for each Region for the table)
                Block    (Blocks within a StoreFile within a Store for each Region for the table)

For a description of what HBase files look like when written to HDFS, see Browsing HDFS for HBase Objects.

70.1. Considerations for Number of Regions

In general, HBase is designed to run with a small (20-200) number of relatively large (5-20Gb) regions per server. The considerations for this are as follows:

70.1.1. Why should I keep my Region count low?

Typically you want to keep your region count low on HBase for numerous reasons. Usually right around 100 regions per RegionServer has yielded the best results. Here are some of the reasons below for keeping region count low:

  1. MSLAB (MemStore-local allocation buffer) requires 2MB per MemStore (that’s 2MB per family per region). 1000 regions that have 2 families each is 3.9GB of heap used, and it’s not even storing data yet. NB: the 2MB value is configurable.

  2. If you fill all the regions at somewhat the same rate, the global memory usage makes it that it forces tiny flushes when you have too many regions which in turn generates compactions. Rewriting the same data tens of times is the last thing you want. An example is filling 1000 regions (with one family) equally and let’s consider a lower bound for global MemStore usage of 5GB (the region server would have a big heap). Once it reaches 5GB it will force flush the biggest region, at that point they should almost all have about 5MB of data so it would flush that amount. 5MB inserted later, it would flush another region that will now have a bit over 5MB of data, and so on. This is currently the main limiting factor for the number of regions; see Number of regions per RS - upper bound for detailed formula.

  3. The master as is is allergic to tons of regions, and will take a lot of time assigning them and moving them around in batches. The reason is that it’s heavy on ZK usage, and it’s not very async at the moment (could really be improved — and has been improved a bunch in 0.96 HBase).

  4. In older versions of HBase (pre-HFile v2, 0.90 and previous), tons of regions on a few RS can cause the store file index to rise, increasing heap usage and potentially creating memory pressure or OOME on the RSs

Another issue is the effect of the number of regions on MapReduce jobs; it is typical to have one mapper per HBase region. Thus, hosting only 5 regions per RS may not be enough to get sufficient number of tasks for a MapReduce job, while 1000 regions will generate far too many tasks.

See Determining region count and size for configuration guidelines.

70.2. Region-RegionServer Assignment

This section describes how Regions are assigned to RegionServers.

70.2.1. Startup

When HBase starts regions are assigned as follows (short version):

  1. The Master invokes the AssignmentManager upon startup.

  2. The AssignmentManager looks at the existing region assignments in hbase:meta.

  3. If the region assignment is still valid (i.e., if the RegionServer is still online) then the assignment is kept.

  4. If the assignment is invalid, then the LoadBalancerFactory is invoked to assign the region. The load balancer (StochasticLoadBalancer by default in HBase 1.0) assign the region to a RegionServer.

  5. hbase:meta is updated with the RegionServer assignment (if needed) and the RegionServer start codes (start time of the RegionServer process) upon region opening by the RegionServer.

70.2.2. Failover

When a RegionServer fails:

  1. The regions immediately become unavailable because the RegionServer is down.

  2. The Master will detect that the RegionServer has failed.

  3. The region assignments will be considered invalid and will be re-assigned just like the startup sequence.

  4. In-flight queries are re-tried, and not lost.

  5. Operations are switched to a new RegionServer within the following amount of time:

    ZooKeeper session timeout + split time + assignment/replay time

70.2.3. Region Load Balancing

Regions can be periodically moved by the LoadBalancer.

70.2.4. Region State Transition

HBase maintains a state for each region and persists the state in hbase:meta. The state of the hbase:meta region itself is persisted in ZooKeeper. You can see the states of regions in transition in the Master web UI. Following is the list of possible region states.

Possible Region States
  • OFFLINE: the region is offline and not opening

  • OPENING: the region is in the process of being opened

  • OPEN: the region is open and the RegionServer has notified the master

  • FAILED_OPEN: the RegionServer failed to open the region

  • CLOSING: the region is in the process of being closed

  • CLOSED: the RegionServer has closed the region and notified the master

  • FAILED_CLOSE: the RegionServer failed to close the region

  • SPLITTING: the RegionServer notified the master that the region is splitting

  • SPLIT: the RegionServer notified the master that the region has finished splitting

  • SPLITTING_NEW: this region is being created by a split which is in progress

  • MERGING: the RegionServer notified the master that this region is being merged with another region

  • MERGED: the RegionServer notified the master that this region has been merged

  • MERGING_NEW: this region is being created by a merge of two regions

region states
Figure 2. Region State Transitions
Graph Legend
  • Brown: Offline state, a special state that can be transient (after closed before opening), terminal (regions of disabled tables), or initial (regions of newly created tables)

  • Palegreen: Online state that regions can serve requests

  • Lightblue: Transient states

  • Red: Failure states that need OPS attention

  • Gold: Terminal states of regions split/merged

  • Grey: Initial states of regions created through split/merge

Transition State Descriptions
  1. The master moves a region from OFFLINE to OPENING state and tries to assign the region to a RegionServer. The RegionServer may or may not have received the open region request. The master retries sending the open region request to the RegionServer until the RPC goes through or the master runs out of retries. After the RegionServer receives the open region request, the RegionServer begins opening the region.

  2. If the master is running out of retries, the master prevents the RegionServer from opening the region by moving the region to CLOSING state and trying to close it, even if the RegionServer is starting to open the region.

  3. After the RegionServer opens the region, it continues to try to notify the master until the master moves the region to OPEN state and notifies the RegionServer. The region is now open.

  4. If the RegionServer cannot open the region, it notifies the master. The master moves the region to CLOSED state and tries to open the region on a different RegionServer.

  5. If the master cannot open the region on any of a certain number of regions, it moves the region to FAILED_OPEN state, and takes no further action until an operator intervenes from the HBase shell, or the server is dead.

  6. The master moves a region from OPEN to CLOSING state. The RegionServer holding the region may or may not have received the close region request. The master retries sending the close request to the server until the RPC goes through or the master runs out of retries.

  7. If the RegionServer is not online, or throws NotServingRegionException, the master moves the region to OFFLINE state and re-assigns it to a different RegionServer.

  8. If the RegionServer is online, but not reachable after the master runs out of retries, the master moves the region to FAILED_CLOSE state and takes no further action until an operator intervenes from the HBase shell, or the server is dead.

  9. If the RegionServer gets the close region request, it closes the region and notifies the master. The master moves the region to CLOSED state and re-assigns it to a different RegionServer.

  10. Before assigning a region, the master moves the region to OFFLINE state automatically if it is in CLOSED state.

  11. When a RegionServer is about to split a region, it notifies the master. The master moves the region to be split from OPEN to SPLITTING state and add the two new regions to be created to the RegionServer. These two regions are in SPLITTING_NEW state initially.

  12. After notifying the master, the RegionServer starts to split the region. Once past the point of no return, the RegionServer notifies the master again so the master can update the hbase:meta table. However, the master does not update the region states until it is notified by the server that the split is done. If the split is successful, the splitting region is moved from SPLITTING to SPLIT state and the two new regions are moved from SPLITTING_NEW to OPEN state.

  13. If the split fails, the splitting region is moved from SPLITTING back to OPEN state, and the two new regions which were created are moved from SPLITTING_NEW to OFFLINE state.

  14. When a RegionServer is about to merge two regions, it notifies the master first. The master moves the two regions to be merged from OPEN to MERGING state, and adds the new region which will hold the contents of the merged regions region to the RegionServer. The new region is in MERGING_NEW state initially.

  15. After notifying the master, the RegionServer starts to merge the two regions. Once past the point of no return, the RegionServer notifies the master again so the master can update the META. However, the master does not update the region states until it is notified by the RegionServer that the merge has completed. If the merge is successful, the two merging regions are moved from MERGING to MERGED state and the new region is moved from MERGING_NEW to OPEN state.

  16. If the merge fails, the two merging regions are moved from MERGING back to OPEN state, and the new region which was created to hold the contents of the merged regions is moved from MERGING_NEW to OFFLINE state.

  17. For regions in FAILED_OPEN or FAILED_CLOSE states, the master tries to close them again when they are reassigned by an operator via HBase Shell.

70.3. Region-RegionServer Locality

Over time, Region-RegionServer locality is achieved via HDFS block replication. The HDFS client does the following by default when choosing locations to write replicas:

  1. First replica is written to local node

  2. Second replica is written to a random node on another rack

  3. Third replica is written on the same rack as the second, but on a different node chosen randomly

  4. Subsequent replicas are written on random nodes on the cluster. See Replica Placement: The First Baby Steps on this page: HDFS Architecture

Thus, HBase eventually achieves locality for a region after a flush or a compaction. In a RegionServer failover situation a RegionServer may be assigned regions with non-local StoreFiles (because none of the replicas are local), however as new data is written in the region, or the table is compacted and StoreFiles are re-written, they will become "local" to the RegionServer.

For more information, see Replica Placement: The First Baby Steps on this page: HDFS Architecture and also Lars George’s blog on HBase and HDFS locality.

70.4. Region Splits

Regions split when they reach a configured threshold. Below we treat the topic in short. For a longer exposition, see Apache HBase Region Splitting and Merging by our Enis Soztutar.

Splits run unaided on the RegionServer; i.e. the Master does not participate. The RegionServer splits a region, offlines the split region and then adds the daughter regions to hbase:meta, opens daughters on the parent’s hosting RegionServer and then reports the split to the Master. See Managed Splitting for how to manually manage splits (and for why you might do this).

70.4.1. Custom Split Policies

You can override the default split policy using a custom RegionSplitPolicy(HBase 0.94+). Typically a custom split policy should extend HBase’s default split policy: IncreasingToUpperBoundRegionSplitPolicy.

The policy can set globally through the HBase configuration or on a per-table basis.

Configuring the Split Policy Globally in hbase-site.xml
Configuring a Split Policy On a Table Using the Java API
HTableDescriptor tableDesc = new HTableDescriptor("test");
tableDesc.setValue(HTableDescriptor.SPLIT_POLICY, ConstantSizeRegionSplitPolicy.class.getName());
tableDesc.addFamily(new HColumnDescriptor(Bytes.toBytes("cf1")));
Configuring the Split Policy On a Table Using HBase Shell
hbase> create 'test', {METHOD => 'table_att', CONFIG => {'SPLIT_POLICY' => 'org.apache.hadoop.hbase.regionserver.ConstantSizeRegionSplitPolicy'}},
{NAME => 'cf1'}

The default split policy can be overwritten using a custom RegionSplitPolicy(HBase 0.94+). Typically a custom split policy should extend HBase’s default split policy: ConstantSizeRegionSplitPolicy.

The policy can be set globally through the HBaseConfiguration used or on a per table basis:

HTableDescriptor myHtd = ...;
myHtd.setValue(HTableDescriptor.SPLIT_POLICY, MyCustomSplitPolicy.class.getName());
The DisabledRegionSplitPolicy policy blocks manual region splitting.

70.5. Manual Region Splitting

It is possible to manually split your table, either at table creation (pre-splitting), or at a later time as an administrative action. You might choose to split your region for one or more of the following reasons. There may be other valid reasons, but the need to manually split your table might also point to problems with your schema design.

Reasons to Manually Split Your Table
  • Your data is sorted by timeseries or another similar algorithm that sorts new data at the end of the table. This means that the Region Server holding the last region is always under load, and the other Region Servers are idle, or mostly idle. See also Monotonically Increasing Row Keys/Timeseries Data.

  • You have developed an unexpected hotspot in one region of your table. For instance, an application which tracks web searches might be inundated by a lot of searches for a celebrity in the event of news about that celebrity. See perf.one.region for more discussion about this particular scenario.

  • After a big increase in the number of RegionServers in your cluster, to get the load spread out quickly.

  • Before a bulk-load which is likely to cause unusual and uneven load across regions.

See Managed Splitting for a discussion about the dangers and possible benefits of managing splitting completely manually.

The DisabledRegionSplitPolicy policy blocks manual region splitting.

70.5.1. Determining Split Points

The goal of splitting your table manually is to improve the chances of balancing the load across the cluster in situations where good rowkey design alone won’t get you there. Keeping that in mind, the way you split your regions is very dependent upon the characteristics of your data. It may be that you already know the best way to split your table. If not, the way you split your table depends on what your keys are like.

Alphanumeric Rowkeys

If your rowkeys start with a letter or number, you can split your table at letter or number boundaries. For instance, the following command creates a table with regions that split at each vowel, so the first region has A-D, the second region has E-H, the third region has I-N, the fourth region has O-V, and the fifth region has U-Z.

Using a Custom Algorithm

The RegionSplitter tool is provided with HBase, and uses a SplitAlgorithm to determine split points for you. As parameters, you give it the algorithm, desired number of regions, and column families. It includes two split algorithms. The first is the HexStringSplit algorithm, which assumes the row keys are hexadecimal strings. The second, UniformSplit, assumes the row keys are random byte arrays. You will probably need to develop your own SplitAlgorithm, using the provided ones as models.

70.6. Online Region Merges

Both Master and RegionServer participate in the event of online region merges. Client sends merge RPC to the master, then the master moves the regions together to the RegionServer where the more heavily loaded region resided. Finally the master sends the merge request to this RegionServer which then runs the merge. Similar to process of region splitting, region merges run as a local transaction on the RegionServer. It offlines the regions and then merges two regions on the file system, atomically delete merging regions from hbase:meta and adds the merged region to hbase:meta, opens the merged region on the RegionServer and reports the merge to the Master.

An example of region merges in the HBase shell

$ hbase> merge_region 'ENCODED_REGIONNAME', 'ENCODED_REGIONNAME', true

It’s an asynchronous operation and call returns immediately without waiting merge completed. Passing true as the optional third parameter will force a merge. Normally only adjacent regions can be merged. The force parameter overrides this behaviour and is for expert use only.

70.7. Store

A Store hosts a MemStore and 0 or more StoreFiles (HFiles). A Store corresponds to a column family for a table for a given region.

70.7.1. MemStore

The MemStore holds in-memory modifications to the Store. Modifications are Cells/KeyValues. When a flush is requested, the current MemStore is moved to a snapshot and is cleared. HBase continues to serve edits from the new MemStore and backing snapshot until the flusher reports that the flush succeeded. At this point, the snapshot is discarded. Note that when the flush happens, MemStores that belong to the same region will all be flushed.

70.7.2. MemStore Flush

A MemStore flush can be triggered under any of the conditions listed below. The minimum flush unit is per region, not at individual MemStore level.

  1. When a MemStore reaches the size specified by hbase.hregion.memstore.flush.size, all MemStores that belong to its region will be flushed out to disk.

  2. When the overall MemStore usage reaches the value specified by hbase.regionserver.global.memstore.upperLimit, MemStores from various regions will be flushed out to disk to reduce overall MemStore usage in a RegionServer.

    The flush order is based on the descending order of a region’s MemStore usage.

    Regions will have their MemStores flushed until the overall MemStore usage drops to or slightly below hbase.regionserver.global.memstore.lowerLimit.

  3. When the number of WAL log entries in a given region server’s WAL reaches the value specified in hbase.regionserver.max.logs, MemStores from various regions will be flushed out to disk to reduce the number of logs in the WAL.

    The flush order is based on time.

    Regions with the oldest MemStores are flushed first until WAL count drops below hbase.regionserver.max.logs.

70.7.3. Scans

  • When a client issues a scan against a table, HBase generates RegionScanner objects, one per region, to serve the scan request.

  • The RegionScanner object contains a list of StoreScanner objects, one per column family.

  • Each StoreScanner object further contains a list of StoreFileScanner objects, corresponding to each StoreFile and HFile of the corresponding column family, and a list of KeyValueScanner objects for the MemStore.

  • The two lists are merged into one, which is sorted in ascending order with the scan object for the MemStore at the end of the list.

  • When a StoreFileScanner object is constructed, it is associated with a MultiVersionConcurrencyControl read point, which is the current memstoreTS, filtering out any new updates beyond the read point.

70.7.4. StoreFile (HFile)

StoreFiles are where your data lives.

HFile Format

The HFile file format is based on the SSTable file described in the BigTable [2006] paper and on Hadoop’s TFile (The unit test suite and the compression harness were taken directly from TFile). Schubert Zhang’s blog post on HFile: A Block-Indexed File Format to Store Sorted Key-Value Pairs makes for a thorough introduction to HBase’s HFile. Matteo Bertozzi has also put up a helpful description, HBase I/O: HFile.

For more information, see the HFile source code. Also see HBase file format with inline blocks (version 2) for information about the HFile v2 format that was included in 0.92.

HFile Tool

To view a textualized version of HFile content, you can use the org.apache.hadoop.hbase.io.hfile.HFile tool. Type the following to see usage:

$ ${HBASE_HOME}/bin/hbase org.apache.hadoop.hbase.io.hfile.HFile

For example, to view the content of the file hdfs://, type the following:

 $ ${HBASE_HOME}/bin/hbase org.apache.hadoop.hbase.io.hfile.HFile -v -f hdfs://

If you leave off the option -v to see just a summary on the HFile. See usage for other things to do with the HFile tool.

In the output of this tool, you might see 'seqid=0' for certain keys in places such as 'Mid-key'/'firstKey'/'lastKey'. These are 'KeyOnlyKeyValue' type instances - meaning their seqid is irrelevant & we just need the keys of these Key-Value instances.
StoreFile Directory Structure on HDFS

For more information of what StoreFiles look like on HDFS with respect to the directory structure, see Browsing HDFS for HBase Objects.

70.7.5. Blocks

StoreFiles are composed of blocks. The blocksize is configured on a per-ColumnFamily basis.

Compression happens at the block level within StoreFiles. For more information on compression, see Compression and Data Block Encoding In HBase.

For more information on blocks, see the HFileBlock source code.

70.7.6. KeyValue

The KeyValue class is the heart of data storage in HBase. KeyValue wraps a byte array and takes offsets and lengths into the passed array which specify where to start interpreting the content as KeyValue.

The KeyValue format inside a byte array is:

  • keylength

  • valuelength

  • key

  • value

The Key is further decomposed as:

  • rowlength

  • row (i.e., the rowkey)

  • columnfamilylength

  • columnfamily

  • columnqualifier

  • timestamp

  • keytype (e.g., Put, Delete, DeleteColumn, DeleteFamily)

KeyValue instances are not split across blocks. For example, if there is an 8 MB KeyValue, even if the block-size is 64kb this KeyValue will be read in as a coherent block. For more information, see the KeyValue source code.


To emphasize the points above, examine what happens with two Puts for two different columns for the same row:

  • Put #1: rowkey=row1, cf:attr1=value1

  • Put #2: rowkey=row1, cf:attr2=value2

Even though these are for the same row, a KeyValue is created for each column:

Key portion for Put #1:

  • rowlength -----------→ 4

  • row -----------------→ row1

  • columnfamilylength --→ 2

  • columnfamily --------→ cf

  • columnqualifier -----→ attr1

  • timestamp -----------→ server time of Put

  • keytype -------------→ Put

Key portion for Put #2:

  • rowlength -----------→ 4

  • row -----------------→ row1

  • columnfamilylength --→ 2

  • columnfamily --------→ cf

  • columnqualifier -----→ attr2

  • timestamp -----------→ server time of Put

  • keytype -------------→ Put

It is critical to understand that the rowkey, ColumnFamily, and column (aka columnqualifier) are embedded within the KeyValue instance. The longer these identifiers are, the bigger the KeyValue is.

70.7.7. Compaction

Ambiguous Terminology
  • A StoreFile is a facade of HFile. In terms of compaction, use of StoreFile seems to have prevailed in the past.

  • A Store is the same thing as a ColumnFamily. StoreFiles are related to a Store, or ColumnFamily.

  • If you want to read more about StoreFiles versus HFiles and Stores versus ColumnFamilies, see HBASE-11316.

When the MemStore reaches a given size (hbase.hregion.memstore.flush.size), it flushes its contents to a StoreFile. The number of StoreFiles in a Store increases over time. Compaction is an operation which reduces the number of StoreFiles in a Store, by merging them together, in order to increase performance on read operations. Compactions can be resource-intensive to perform, and can either help or hinder performance depending on many factors.

Compactions fall into two categories: minor and major. Minor and major compactions differ in the following ways.

Minor compactions usually select a small number of small, adjacent StoreFiles and rewrite them as a single StoreFile. Minor compactions do not drop (filter out) deletes or expired versions, because of potential side effects. See Compaction and Deletions and Compaction and Versions for information on how deletes and versions are handled in relation to compactions. The end result of a minor compaction is fewer, larger StoreFiles for a given Store.

The end result of a major compaction is a single StoreFile per Store. Major compactions also process delete markers and max versions. See Compaction and Deletions and Compaction and Versions for information on how deletes and versions are handled in relation to compactions.

Compaction and Deletions

When an explicit deletion occurs in HBase, the data is not actually deleted. Instead, a tombstone marker is written. The tombstone marker prevents the data from being returned with queries. During a major compaction, the data is actually deleted, and the tombstone marker is removed from the StoreFile. If the deletion happens because of an expired TTL, no tombstone is created. Instead, the expired data is filtered out and is not written back to the compacted StoreFile.

Compaction and Versions

When you create a Column Family, you can specify the maximum number of versions to keep, by specifying HColumnDescriptor.setMaxVersions(int versions). The default value is 3. If more versions than the specified maximum exist, the excess versions are filtered out and not written back to the compacted StoreFile.

Major Compactions Can Impact Query Results

In some situations, older versions can be inadvertently resurrected if a newer version is explicitly deleted. See Major compactions change query results for a more in-depth explanation. This situation is only possible before the compaction finishes.

In theory, major compactions improve performance. However, on a highly loaded system, major compactions can require an inappropriate number of resources and adversely affect performance. In a default configuration, major compactions are scheduled automatically to run once in a 7-day period. This is sometimes inappropriate for systems in production. You can manage major compactions manually. See Managed Compactions.

Compactions do not perform region merges. See Merge for more information on region merging.

Compaction Policy - HBase 0.96.x and newer

Compacting large StoreFiles, or too many StoreFiles at once, can cause more IO load than your cluster is able to handle without causing performance problems. The method by which HBase selects which StoreFiles to include in a compaction (and whether the compaction is a minor or major compaction) is called the compaction policy.

Prior to HBase 0.96.x, there was only one compaction policy. That original compaction policy is still available as RatioBasedCompactionPolicy. The new compaction default policy, called ExploringCompactionPolicy, was subsequently backported to HBase 0.94 and HBase 0.95, and is the default in HBase 0.96 and newer. It was implemented in HBASE-7842. In short, ExploringCompactionPolicy attempts to select the best possible set of StoreFiles to compact with the least amount of work, while the RatioBasedCompactionPolicy selects the first set that meets the criteria.

Regardless of the compaction policy used, file selection is controlled by several configurable parameters and happens in a multi-step approach. These parameters will be explained in context, and then will be given in a table which shows their descriptions, defaults, and implications of changing them.

Being Stuck

When the MemStore gets too large, it needs to flush its contents to a StoreFile. However, a Store can only have hbase.hstore.blockingStoreFiles files, so the MemStore needs to wait for the number of StoreFiles to be reduced by one or more compactions. However, if the MemStore grows larger than hbase.hregion.memstore.flush.size, it is not able to flush its contents to a StoreFile. If the MemStore is too large and the number of StoreFiles is also too high, the algorithm is said to be "stuck". The compaction algorithm checks for this "stuck" situation and provides mechanisms to alleviate it.

The ExploringCompactionPolicy Algorithm

The ExploringCompactionPolicy algorithm considers each possible set of adjacent StoreFiles before choosing the set where compaction will have the most benefit.

One situation where the ExploringCompactionPolicy works especially well is when you are bulk-loading data and the bulk loads create larger StoreFiles than the StoreFiles which are holding data older than the bulk-loaded data. This can "trick" HBase into choosing to perform a major compaction each time a compaction is needed, and cause a lot of extra overhead. With the ExploringCompactionPolicy, major compactions happen much less frequently because minor compactions are more efficient.

In general, ExploringCompactionPolicy is the right choice for most situations, and thus is the default compaction policy. You can also use ExploringCompactionPolicy along with Experimental: Stripe Compactions.

The logic of this policy can be examined in hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/compactions/ExploringCompactionPolicy.java. The following is a walk-through of the logic of the ExploringCompactionPolicy.

  1. Make a list of all existing StoreFiles in the Store. The rest of the algorithm filters this list to come up with the subset of HFiles which will be chosen for compaction.

  2. If this was a user-requested compaction, attempt to perform the requested compaction type, regardless of what would normally be chosen. Note that even if the user requests a major compaction, it may not be possible to perform a major compaction. This may be because not all StoreFiles in the Column Family are available to compact or because there are too many Stores in the Column Family.

  3. Some StoreFiles are automatically excluded from consideration. These include:

    • StoreFiles that are larger than hbase.hstore.compaction.max.size

    • StoreFiles that were created by a bulk-load operation which explicitly excluded compaction. You may decide to exclude StoreFiles resulting from bulk loads, from compaction. To do this, specify the hbase.mapreduce.hfileoutputformat.compaction.exclude parameter during the bulk load operation.

  4. Iterate through the list from step 1, and make a list of all potential sets of StoreFiles to compact together. A potential set is a grouping of hbase.hstore.compaction.min contiguous StoreFiles in the list. For each set, perform some sanity-checking and figure out whether this is the best compaction that could be done:

    • If the number of StoreFiles in this set (not the size of the StoreFiles) is fewer than hbase.hstore.compaction.min or more than hbase.hstore.compaction.max, take it out of consideration.

    • Compare the size of this set of StoreFiles with the size of the smallest possible compaction that has been found in the list so far. If the size of this set of StoreFiles represents the smallest compaction that could be done, store it to be used as a fall-back if the algorithm is "stuck" and no StoreFiles would otherwise be chosen. See Being Stuck.

    • Do size-based sanity checks against each StoreFile in this set of StoreFiles.

      • If the size of this StoreFile is larger than hbase.hstore.compaction.max.size, take it out of consideration.

      • If the size is greater than or equal to hbase.hstore.compaction.min.size, sanity-check it against the file-based ratio to see whether it is too large to be considered.

        The sanity-checking is successful if:

      • There is only one StoreFile in this set, or

      • For each StoreFile, its size multiplied by hbase.hstore.compaction.ratio (or hbase.hstore.compaction.ratio.offpeak if off-peak hours are configured and it is during off-peak hours) is less than the sum of the sizes of the other HFiles in the set.

  5. If this set of StoreFiles is still in consideration, compare it to the previously-selected best compaction. If it is better, replace the previously-selected best compaction with this one.

  6. When the entire list of potential compactions has been processed, perform the best compaction that was found. If no StoreFiles were selected for compaction, but there are multiple StoreFiles, assume the algorithm is stuck (see Being Stuck) and if so, perform the smallest compaction that was found in step 3.

RatioBasedCompactionPolicy Algorithm

The RatioBasedCompactionPolicy was the only compaction policy prior to HBase 0.96, though ExploringCompactionPolicy has now been backported to HBase 0.94 and 0.95. To use the RatioBasedCompactionPolicy rather than the ExploringCompactionPolicy, set hbase.hstore.defaultengine.compactionpolicy.class to RatioBasedCompactionPolicy in the hbase-site.xml file. To switch back to the ExploringCompactionPolicy, remove the setting from the hbase-site.xml.

The following section walks you through the algorithm used to select StoreFiles for compaction in the RatioBasedCompactionPolicy.

  1. The first phase is to create a list of all candidates for compaction. A list is created of all StoreFiles not already in the compaction queue, and all StoreFiles newer than the newest file that is currently being compacted. This list of StoreFiles is ordered by the sequence ID. The sequence ID is generated when a Put is appended to the write-ahead log (WAL), and is stored in the metadata of the HFile.

  2. Check to see if the algorithm is stuck (see Being Stuck, and if so, a major compaction is forced. This is a key area where The ExploringCompactionPolicy Algorithm is often a better choice than the RatioBasedCompactionPolicy.

  3. If the compaction was user-requested, try to perform the type of compaction that was requested. Note that a major compaction may not be possible if all HFiles are not available for compaction or if too many StoreFiles exist (more than hbase.hstore.compaction.max).

  4. Some StoreFiles are automatically excluded from consideration. These include:

    • StoreFiles that are larger than hbase.hstore.compaction.max.size

    • StoreFiles that were created by a bulk-load operation which explicitly excluded compaction. You may decide to exclude StoreFiles resulting from bulk loads, from compaction. To do this, specify the hbase.mapreduce.hfileoutputformat.compaction.exclude parameter during the bulk load operation.

  5. The maximum number of StoreFiles allowed in a major compaction is controlled by the hbase.hstore.compaction.max parameter. If the list contains more than this number of StoreFiles, a minor compaction is performed even if a major compaction would otherwise have been done. However, a user-requested major compaction still occurs even if there are more than hbase.hstore.compaction.max StoreFiles to compact.

  6. If the list contains fewer than hbase.hstore.compaction.min StoreFiles to compact, a minor compaction is aborted. Note that a major compaction can be performed on a single HFile. Its function is to remove deletes and expired versions, and reset locality on the StoreFile.

  7. The value of the hbase.hstore.compaction.ratio parameter is multiplied by the sum of StoreFiles smaller than a given file, to determine whether that StoreFile is selected for compaction during a minor compaction. For instance, if hbase.hstore.compaction.ratio is 1.2, FileX is 5MB, FileY is 2MB, and FileZ is 3MB:

    5 <= 1.2 x (2 + 3)            or            5 <= 6

    In this scenario, FileX is eligible for minor compaction. If FileX were 7MB, it would not be eligible for minor compaction. This ratio favors smaller StoreFile. You can configure a different ratio for use in off-peak hours, using the parameter hbase.hstore.compaction.ratio.offpeak, if you also configure hbase.offpeak.start.hour and hbase.offpeak.end.hour.

  8. If the last major compaction was too long ago and there is more than one StoreFile to be compacted, a major compaction is run, even if it would otherwise have been minor. By default, the maximum time between major compactions is 7 days, plus or minus a 4.8 hour period, and determined randomly within those parameters. Prior to HBase 0.96, the major compaction period was 24 hours. See hbase.hregion.majorcompaction in the table below to tune or disable time-based major compactions.

Parameters Used by Compaction Algorithm

This table contains the main configuration parameters for compaction. This list is not exhaustive. To tune these parameters from the defaults, edit the hbase-default.xml file. For a full list of all configuration parameters available, see config.files


The minimum number of StoreFiles which must be eligible for compaction before compaction can run. The goal of tuning hbase.hstore.compaction.min is to avoid ending up with too many tiny StoreFiles to compact. Setting this value to 2 would cause a minor compaction each time you have two StoreFiles in a Store, and this is probably not appropriate. If you set this value too high, all the other values will need to be adjusted accordingly. For most cases, the default value is appropriate. In previous versions of HBase, the parameter hbase.hstore.compaction.min was called hbase.hstore.compactionThreshold.

Default: 3


The maximum number of StoreFiles which will be selected for a single minor compaction, regardless of the number of eligible StoreFiles. Effectively, the value of hbase.hstore.compaction.max controls the length of time it takes a single compaction to complete. Setting it larger means that more StoreFiles are included in a compaction. For most cases, the default value is appropriate.

Default: 10


A StoreFile smaller than this size will always be eligible for minor compaction. StoreFiles this size or larger are evaluated by hbase.hstore.compaction.ratio to determine if they are eligible. Because this limit represents the "automatic include" limit for all StoreFiles smaller than this value, this value may need to be reduced in write-heavy environments where many files in the 1-2 MB range are being flushed, because every StoreFile will be targeted for compaction and the resulting StoreFiles may still be under the minimum size and require further compaction. If this parameter is lowered, the ratio check is triggered more quickly. This addressed some issues seen in earlier versions of HBase but changing this parameter is no longer necessary in most situations.

Default:128 MB


A StoreFile larger than this size will be excluded from compaction. The effect of raising hbase.hstore.compaction.max.size is fewer, larger StoreFiles that do not get compacted often. If you feel that compaction is happening too often without much benefit, you can try raising this value.

Default: Long.MAX_VALUE


For minor compaction, this ratio is used to determine whether a given StoreFile which is larger than hbase.hstore.compaction.min.size is eligible for compaction. Its effect is to limit compaction of large StoreFile. The value of hbase.hstore.compaction.ratio is expressed as a floating-point decimal.

  • A large ratio, such as 10, will produce a single giant StoreFile. Conversely, a value of .25, will produce behavior similar to the BigTable compaction algorithm, producing four StoreFiles.

  • A moderate value of between 1.0 and 1.4 is recommended. When tuning this value, you are balancing write costs with read costs. Raising the value (to something like 1.4) will have more write costs, because you will compact larger StoreFiles. However, during reads, HBase will need to seek through fewer StoreFiles to accomplish the read. Consider this approach if you cannot take advantage of Bloom Filters.

  • Alternatively, you can lower this value to something like 1.0 to reduce the background cost of writes, and use to limit the number of StoreFiles touched during reads. For most cases, the default value is appropriate.

    Default: 1.2F


The compaction ratio used during off-peak compactions, if off-peak hours are also configured (see below). Expressed as a floating-point decimal. This allows for more aggressive (or less aggressive, if you set it lower than hbase.hstore.compaction.ratio) compaction during a set time period. Ignored if off-peak is disabled (default). This works the same as hbase.hstore.compaction.ratio.

Default: 5.0F


The start of off-peak hours, expressed as an integer between 0 and 23, inclusive. Set to -1 to disable off-peak.

Default: -1 (disabled)


The end of off-peak hours, expressed as an integer between 0 and 23, inclusive. Set to -1 to disable off-peak.

Default: -1 (disabled)


There are two different thread pools for compactions, one for large compactions and the other for small compactions. This helps to keep compaction of lean tables (such as hbase:meta) fast. If a compaction is larger than this threshold, it goes into the large compaction pool. In most cases, the default value is appropriate.

Default: 2 x hbase.hstore.compaction.max x hbase.hregion.memstore.flush.size (which defaults to 128)


Time between major compactions, expressed in milliseconds. Set to 0 to disable time-based automatic major compactions. User-requested and size-based major compactions will still run. This value is multiplied by hbase.hregion.majorcompaction.jitter to cause compaction to start at a somewhat-random time during a given window of time.

Default: 7 days (604800000 milliseconds)


A multiplier applied to hbase.hregion.majorcompaction to cause compaction to occur a given amount of time either side of hbase.hregion.majorcompaction. The smaller the number, the closer the compactions will happen to the hbase.hregion.majorcompaction interval. Expressed as a floating-point decimal.

Default: .50F

Compaction File Selection
Legacy Information

This section has been preserved for historical reasons and refers to the way compaction worked prior to HBase 0.96.x. You can still use this behavior if you enable RatioBasedCompactionPolicy Algorithm. For information on the way that compactions work in HBase 0.96.x and later, see Compaction.

To understand the core algorithm for StoreFile selection, there is some ASCII-art in the Store source code that will serve as useful reference.

It has been copied below:

/* normal skew:
 *         older ----> newer
 *     _
 *    | |   _
 *    | |  | |   _
 *  --|-|- |-|- |-|---_-------_-------  minCompactSize
 *    | |  | |  | |  | |  _  | |
 *    | |  | |  | |  | | | | | |
 *    | |  | |  | |  | | | | | |
Important knobs:
  • hbase.hstore.compaction.ratio Ratio used in compaction file selection algorithm (default 1.2f).

  • hbase.hstore.compaction.min (in HBase v 0.90 this is called hbase.hstore.compactionThreshold) (files) Minimum number of StoreFiles per Store to be selected for a compaction to occur (default 2).

  • hbase.hstore.compaction.max (files) Maximum number of StoreFiles to compact per minor compaction (default 10).

  • hbase.hstore.compaction.min.size (bytes) Any StoreFile smaller than this setting with automatically be a candidate for compaction. Defaults to hbase.hregion.memstore.flush.size (128 mb).

  • hbase.hstore.compaction.max.size (.92) (bytes) Any StoreFile larger than this setting with automatically be excluded from compaction (default Long.MAX_VALUE).

The minor compaction StoreFile selection logic is size based, and selects a file for compaction when the file ⇐ sum(smaller_files) * hbase.hstore.compaction.ratio.

Minor Compaction File Selection - Example #1 (Basic Example)

This example mirrors an example from the unit test TestCompactSelection.

  • hbase.hstore.compaction.ratio = 1.0f

  • hbase.hstore.compaction.min = 3 (files)

  • hbase.hstore.compaction.max = 5 (files)

  • hbase.hstore.compaction.min.size = 10 (bytes)

  • hbase.hstore.compaction.max.size = 1000 (bytes)

The following StoreFiles exist: 100, 50, 23, 12, and 12 bytes apiece (oldest to newest). With the above parameters, the files that would be selected for minor compaction are 23, 12, and 12.


  • 100 → No, because sum(50, 23, 12, 12) * 1.0 = 97.

  • 50 → No, because sum(23, 12, 12) * 1.0 = 47.

  • 23 → Yes, because sum(12, 12) * 1.0 = 24.

  • 12 → Yes, because the previous file has been included, and because this does not exceed the max-file limit of 5

  • 12 → Yes, because the previous file had been included, and because this does not exceed the max-file limit of 5.

Minor Compaction File Selection - Example #2 (Not Enough Files ToCompact)

This example mirrors an example from the unit test TestCompactSelection.

  • hbase.hstore.compaction.ratio = 1.0f

  • hbase.hstore.compaction.min = 3 (files)

  • hbase.hstore.compaction.max = 5 (files)

  • hbase.hstore.compaction.min.size = 10 (bytes)

  • hbase.hstore.compaction.max.size = 1000 (bytes)

The following StoreFiles exist: 100, 25, 12, and 12 bytes apiece (oldest to newest). With the above parameters, no compaction will be started.


  • 100 → No, because sum(25, 12, 12) * 1.0 = 47

  • 25 → No, because sum(12, 12) * 1.0 = 24

  • 12 → No. Candidate because sum(12) * 1.0 = 12, there are only 2 files to compact and that is less than the threshold of 3

  • 12 → No. Candidate because the previous StoreFile was, but there are not enough files to compact

Minor Compaction File Selection - Example #3 (Limiting Files To Compact)

This example mirrors an example from the unit test TestCompactSelection.

  • hbase.hstore.compaction.ratio = 1.0f

  • hbase.hstore.compaction.min = 3 (files)

  • hbase.hstore.compaction.max = 5 (files)

  • hbase.hstore.compaction.min.size = 10 (bytes)

  • hbase.hstore.compaction.max.size = 1000 (bytes)

The following StoreFiles exist: 7, 6, 5, 4, 3, 2, and 1 bytes apiece (oldest to newest). With the above parameters, the files that would be selected for minor compaction are 7, 6, 5, 4, 3.


  • 7 → Yes, because sum(6, 5, 4, 3, 2, 1) * 1.0 = 21. Also, 7 is less than the min-size

  • 6 → Yes, because sum(5, 4, 3, 2, 1) * 1.0 = 15. Also, 6 is less than the min-size.

  • 5 → Yes, because sum(4, 3, 2, 1) * 1.0 = 10. Also, 5 is less than the min-size.

  • 4 → Yes, because sum(3, 2, 1) * 1.0 = 6. Also, 4 is less than the min-size.

  • 3 → Yes, because sum(2, 1) * 1.0 = 3. Also, 3 is less than the min-size.

  • 2 → No. Candidate because previous file was selected and 2 is less than the min-size, but the max-number of files to compact has been reached.

  • 1 → No. Candidate because previous file was selected and 1 is less than the min-size, but max-number of files to compact has been reached.

Impact of Key Configuration Options
This information is now included in the configuration parameter table in Parameters Used by Compaction Algorithm.
Experimental: Stripe Compactions

Stripe compactions is an experimental feature added in HBase 0.98 which aims to improve compactions for large regions or non-uniformly distributed row keys. In order to achieve smaller and/or more granular compactions, the StoreFiles within a region are maintained separately for several row-key sub-ranges, or "stripes", of the region. The stripes are transparent to the rest of HBase, so other operations on the HFiles or data work without modification.

Stripe compactions change the HFile layout, creating sub-regions within regions. These sub-regions are easier to compact, and should result in fewer major compactions. This approach alleviates some of the challenges of larger regions.

Stripe compaction is fully compatible with Compaction and works in conjunction with either the ExploringCompactionPolicy or RatioBasedCompactionPolicy. It can be enabled for existing tables, and the table will continue to operate normally if it is disabled later.

When To Use Stripe Compactions

Consider using stripe compaction if you have either of the following:

  • Large regions. You can get the positive effects of smaller regions without additional overhead for MemStore and region management overhead.

  • Non-uniform keys, such as time dimension in a key. Only the stripes receiving the new keys will need to compact. Old data will not compact as often, if at all

Performance Improvements

Performance testing has shown that the performance of reads improves somewhat, and variability of performance of reads and writes is greatly reduced. An overall long-term performance improvement is seen on large non-uniform-row key regions, such as a hash-prefixed timestamp key. These performance gains are the most dramatic on a table which is already large. It is possible that the performance improvement might extend to region splits.

Enabling Stripe Compaction

You can enable stripe compaction for a table or a column family, by setting its hbase.hstore.engine.class to org.apache.hadoop.hbase.regionserver.StripeStoreEngine. You also need to set the hbase.hstore.blockingStoreFiles to a high number, such as 100 (rather than the default value of 10).

Procedure: Enable Stripe Compaction
  1. Run one of following commands in the HBase shell. Replace the table name orders_table with the name of your table.

    alter 'orders_table', CONFIGURATION => {'hbase.hstore.engine.class' => 'org.apache.hadoop.hbase.regionserver.StripeStoreEngine', 'hbase.hstore.blockingStoreFiles' => '100'}
    alter 'orders_table', {NAME => 'blobs_cf', CONFIGURATION => {'hbase.hstore.engine.class' => 'org.apache.hadoop.hbase.regionserver.StripeStoreEngine', 'hbase.hstore.blockingStoreFiles' => '100'}}
    create 'orders_table', 'blobs_cf', CONFIGURATION => {'hbase.hstore.engine.class' => 'org.apache.hadoop.hbase.regionserver.StripeStoreEngine', 'hbase.hstore.blockingStoreFiles' => '100'}
  2. Configure other options if needed. See Configuring Stripe Compaction for more information.

  3. Enable the table.

Procedure: Disable Stripe Compaction
  1. Set the hbase.hstore.engine.class option to either nil or org.apache.hadoop.hbase.regionserver.DefaultStoreEngine. Either option has the same effect.

    alter 'orders_table', CONFIGURATION => {'hbase.hstore.engine.class' => 'rg.apache.hadoop.hbase.regionserver.DefaultStoreEngine'}
  2. Enable the table.

When you enable a large table after changing the store engine either way, a major compaction will likely be performed on most regions. This is not necessary on new tables.

Configuring Stripe Compaction

Each of the settings for stripe compaction should be configured at the table or column family level. If you use HBase shell, the general command pattern is as follows:

alter 'orders_table', CONFIGURATION => {'key' => 'value', ..., 'key' => 'value'}}
Region and stripe sizing

You can configure your stripe sizing based upon your region sizing. By default, your new regions will start with one stripe. On the next compaction after the stripe has grown too large (16 x MemStore flushes size), it is split into two stripes. Stripe splitting continues as the region grows, until the region is large enough to split.

You can improve this pattern for your own data. A good rule is to aim for a stripe size of at least 1 GB, and about 8-12 stripes for uniform row keys. For example, if your regions are 30 GB, 12 x 2.5 GB stripes might be a good starting point.

Table 9. Stripe Sizing Settings
Setting Notes


The number of stripes to create when stripe compaction is enabled. You can use it as follows:

  • For relatively uniform row keys, if you know the approximate target number of stripes from the above, you can avoid some splitting overhead by starting with several stripes (2, 5, 10…​). If the early data is not representative of overall row key distribution, this will not be as efficient.

  • For existing tables with a large amount of data, this setting will effectively pre-split your stripes.

  • For keys such as hash-prefixed sequential keys, with more than one hash prefix per region, pre-splitting may make sense.


The maximum size a stripe grows before splitting. Use this in conjunction with hbase.store.stripe.splitPartCount to control the target stripe size (sizeToSplit = splitPartsCount * target stripe size), according to the above sizing considerations.


The number of new stripes to create when splitting a stripe. The default is 2, which is appropriate for most cases. For non-uniform row keys, you can experiment with increasing the number to 3 or 4, to isolate the arriving updates into narrower slice of the region without additional splits being required.

MemStore Size Settings

By default, the flush creates several files from one MemStore, according to existing stripe boundaries and row keys to flush. This approach minimizes write amplification, but can be undesirable if the MemStore is small and there are many stripes, because the files will be too small.

In this type of situation, you can set hbase.store.stripe.compaction.flushToL0 to true. This will cause a MemStore flush to create a single file instead. When at least hbase.store.stripe.compaction.minFilesL0 such files (by default, 4) accumulate, they will be compacted into striped files.

Normal Compaction Configuration and Stripe Compaction

All the settings that apply to normal compactions (see Parameters Used by Compaction Algorithm) apply to stripe compactions. The exceptions are the minimum and maximum number of files, which are set to higher values by default because the files in stripes are smaller. To control these for stripe compactions, use hbase.store.stripe.compaction.minFiles and hbase.store.stripe.compaction.maxFiles, rather than hbase.hstore.compaction.min and hbase.hstore.compaction.max.

71. Bulk Loading

71.1. Overview

HBase includes several methods of loading data into tables. The most straightforward method is to either use the TableOutputFormat class from a MapReduce job, or use the normal client APIs; however, these are not always the most efficient methods.

The bulk load feature uses a MapReduce job to output table data in HBase’s internal data format, and then directly loads the generated StoreFiles into a running cluster. Using bulk load will use less CPU and network resources than simply using the HBase API.

71.2. Bulk Load Limitations

As bulk loading bypasses the write path, the WAL doesn’t get written to as part of the process. Replication works by reading the WAL files so it won’t see the bulk loaded data – and the same goes for the edits that use Put.setDurability(SKIP_WAL). One way to handle that is to ship the raw files or the HFiles to the other cluster and do the other processing there.

71.3. Bulk Load Architecture

The HBase bulk load process consists of two main steps.

71.3.1. Preparing data via a MapReduce job

The first step of a bulk load is to generate HBase data files (StoreFiles) from a MapReduce job using HFileOutputFormat2. This output format writes out data in HBase’s internal storage format so that they can be later loaded very efficiently into the cluster.

In order to function efficiently, HFileOutputFormat2 must be configured such that each output HFile fits within a single region. In order to do this, jobs whose output will be bulk loaded into HBase use Hadoop’s TotalOrderPartitioner class to partition the map output into disjoint ranges of the key space, corresponding to the key ranges of the regions in the table.

HFileOutputFormat2 includes a convenience function, configureIncrementalLoad(), which automatically sets up a TotalOrderPartitioner based on the current region boundaries of a table.

71.3.2. Completing the data load

After a data import has been prepared, either by using the importtsv tool with the “importtsv.bulk.output” option or by some other MapReduce job using the HFileOutputFormat, the completebulkload tool is used to import the data into the running cluster. This command line tool iterates through the prepared data files, and for each one determines the region the file belongs to. It then contacts the appropriate RegionServer which adopts the HFile, moving it into its storage directory and making the data available to clients.

If the region boundaries have changed during the course of bulk load preparation, or between the preparation and completion steps, the completebulkload utility will automatically split the data files into pieces corresponding to the new boundaries. This process is not optimally efficient, so users should take care to minimize the delay between preparing a bulk load and importing it into the cluster, especially if other clients are simultaneously loading data through other means.

$ hadoop jar hbase-server-VERSION.jar completebulkload [-c /path/to/hbase/config/hbase-site.xml] /user/todd/myoutput mytable

The -c config-file option can be used to specify a file containing the appropriate hbase parameters (e.g., hbase-site.xml) if not supplied already on the CLASSPATH (In addition, the CLASSPATH must contain the directory that has the zookeeper configuration file if zookeeper is NOT managed by HBase).

If the target table does not already exist in HBase, this tool will create the table automatically.

71.4. See Also

For more information about the referenced utilities, see ImportTsv and CompleteBulkLoad.

See How-to: Use HBase Bulk Loading, and Why for a recent blog on current state of bulk loading.

71.5. Advanced Usage

Although the importtsv tool is useful in many cases, advanced users may want to generate data programmatically, or import data from other formats. To get started doing so, dig into ImportTsv.java and check the JavaDoc for HFileOutputFormat.

The import step of the bulk load can also be done programmatically. See the LoadIncrementalHFiles class for more information.

72. HDFS

As HBase runs on HDFS (and each StoreFile is written as a file on HDFS), it is important to have an understanding of the HDFS Architecture especially in terms of how it stores files, handles failovers, and replicates blocks.

See the Hadoop documentation on HDFS Architecture for more information.

72.1. NameNode

The NameNode is responsible for maintaining the filesystem metadata. See the above HDFS Architecture link for more information.

72.2. DataNode

The DataNodes are responsible for storing HDFS blocks. See the above HDFS Architecture link for more information.

73. Timeline-consistent High Available Reads

73.1. Introduction

HBase, architecturally, always had the strong consistency guarantee from the start. All reads and writes are routed through a single region server, which guarantees that all writes happen in an order, and all reads are seeing the most recent committed data.

However, because of this single homing of the reads to a single location, if the server becomes unavailable, the regions of the table that were hosted in the region server become unavailable for some time. There are three phases in the region recovery process - detection, assignment, and recovery. Of these, the detection is usually the longest and is presently in the order of 20-30 seconds depending on the ZooKeeper session timeout. During this time and before the recovery is complete, the clients will not be able to read the region data.

However, for some use cases, either the data may be read-only, or doing reads against some stale data is acceptable. With timeline-consistent high available reads, HBase can be used for these kind of latency-sensitive use cases where the application can expect to have a time bound on the read completion.

For achieving high availability for reads, HBase provides a feature called region replication. In this model, for each region of a table, there will be multiple replicas that are opened in different RegionServers. By default, the region replication is set to 1, so only a single region replica is deployed and there will not be any changes from the original model. If region replication is set to 2 or more, then the master will assign replicas of the regions of the table. The Load Balancer ensures that the region replicas are not co-hosted in the same region servers and also in the same rack (if possible).

All of the replicas for a single region will have a unique replica_id, starting from 0. The region replica having replica_id==0 is called the primary region, and the others secondary regions or secondaries. Only the primary can accept writes from the client, and the primary will always contain the latest changes. Since all writes still have to go through the primary region, the writes are not highly-available (meaning they might block for some time if the region becomes unavailable).

73.2. Timeline Consistency

With this feature, HBase introduces a Consistency definition, which can be provided per read operation (get or scan).

public enum Consistency {

Consistency.STRONG is the default consistency model provided by HBase. In case the table has region replication = 1, or in a table with region replicas but the reads are done with this consistency, the read is always performed by the primary regions, so that there will not be any change from the previous behaviour, and the client always observes the latest data.

In case a read is performed with Consistency.TIMELINE, then the read RPC will be sent to the primary region server first. After a short interval (hbase.client.primaryCallTimeout.get, 10ms by default), parallel RPC for secondary region replicas will also be sent if the primary does not respond back. After this, the result is returned from whichever RPC is finished first. If the response came back from the primary region replica, we can always know that the data is latest. For this Result.isStale() API has been added to inspect the staleness. If the result is from a secondary region, then Result.isStale() will be set to true. The user can then inspect this field to possibly reason about the data.

In terms of semantics, TIMELINE consistency as implemented by HBase differs from pure eventual consistency in these respects:

  • Single homed and ordered updates: Region replication or not, on the write side, there is still only 1 defined replica (primary) which can accept writes. This replica is responsible for ordering the edits and preventing conflicts. This guarantees that two different writes are not committed at the same time by different replicas and the data diverges. With this, there is no need to do read-repair or last-timestamp-wins kind of conflict resolution.

  • The secondaries also apply the edits in the order that the primary committed them. This way the secondaries will contain a snapshot of the primaries data at any point in time. This is similar to RDBMS replications and even HBase’s own multi-datacenter replication, however in a single cluster.

  • On the read side, the client can detect whether the read is coming from up-to-date data or is stale data. Also, the client can issue reads with different consistency requirements on a per-operation basis to ensure its own semantic guarantees.

  • The client can still observe edits out-of-order, and can go back in time, if it observes reads from one secondary replica first, then another secondary replica. There is no stickiness to region replicas or a transaction-id based guarantee. If required, this can be implemented later though.

Timeline Consistency
Figure 3. Timeline Consistency

To better understand the TIMELINE semantics, let’s look at the above diagram. Let’s say that there are two clients, and the first one writes x=1 at first, then x=2 and x=3 later. As above, all writes are handled by the primary region replica. The writes are saved in the write ahead log (WAL), and replicated to the other replicas asynchronously. In the above diagram, notice that replica_id=1 received 2 updates, and its data shows that x=2, while the replica_id=2 only received a single update, and its data shows that x=1.

If client1 reads with STRONG consistency, it will only talk with the replica_id=0, and thus is guaranteed to observe the latest value of x=3. In case of a client issuing TIMELINE consistency reads, the RPC will go to all replicas (after primary timeout) and the result from the first response will be returned back. Thus the client can see either 1, 2 or 3 as the value of x. Let’s say that the primary region has failed and log replication cannot continue for some time. If the client does multiple reads with TIMELINE consistency, she can observe x=2 first, then x=1, and so on.

73.3. Tradeoffs

Having secondary regions hosted for read availability comes with some tradeoffs which should be carefully evaluated per use case. Following are advantages and disadvantages.

  • High availability for read-only tables

  • High availability for stale reads

  • Ability to do very low latency reads with very high percentile (99.9%+) latencies for stale reads

  • Double / Triple MemStore usage (depending on region replication count) for tables with region replication > 1

  • Increased block cache usage

  • Extra network traffic for log replication

  • Extra backup RPCs for replicas

To serve the region data from multiple replicas, HBase opens the regions in secondary mode in the region servers. The regions opened in secondary mode will share the same data files with the primary region replica, however each secondary region replica will have its own MemStore to keep the unflushed data (only primary region can do flushes). Also to serve reads from secondary regions, the blocks of data files may be also cached in the block caches for the secondary regions.

73.4. Where is the code

This feature is delivered in two phases, Phase 1 and 2. The first phase is done in time for HBase-1.0.0 release. Meaning that using HBase-1.0.x, you can use all the features that are marked for Phase 1. Phase 2 is committed in HBase-1.1.0, meaning all HBase versions after 1.1.0 should contain Phase 2 items.

73.5. Propagating writes to region replicas

As discussed above writes only go to the primary region replica. For propagating the writes from the primary region replica to the secondaries, there are two different mechanisms. For read-only tables, you do not need to use any of the following methods. Disabling and enabling the table should make the data available in all region replicas. For mutable tables, you have to use only one of the following mechanisms: storefile refresher, or async wal replication. The latter is recommended.

73.5.1. StoreFile Refresher

The first mechanism is store file refresher which is introduced in HBase-1.0+. Store file refresher is a thread per region server, which runs periodically, and does a refresh operation for the store files of the primary region for the secondary region replicas. If enabled, the refresher will ensure that the secondary region replicas see the new flushed, compacted or bulk loaded files from the primary region in a timely manner. However, this means that only flushed data can be read back from the secondary region replicas, and after the refresher is run, making the secondaries lag behind the primary for an a longer time.

For turning this feature on, you should configure hbase.regionserver.storefile.refresh.period to a non-zero value. See Configuration section below.

73.5.2. Asnyc WAL replication

The second mechanism for propagation of writes to secondaries is done via “Async WAL Replication” feature and is only available in HBase-1.1+. This works similarly to HBase’s multi-datacenter replication, but instead the data from a region is replicated to the secondary regions. Each secondary replica always receives and observes the writes in the same order that the primary region committed them. In some sense, this design can be thought of as “in-cluster replication”, where instead of replicating to a different datacenter, the data goes to secondary regions to keep secondary region’s in-memory state up to date. The data files are shared between the primary region and the other replicas, so that there is no extra storage overhead. However, the secondary regions will have recent non-flushed data in their memstores, which increases the memory overhead. The primary region writes flush, compaction, and bulk load events to its WAL as well, which are also replicated through wal replication to secondaries. When they observe the flush/compaction or bulk load event, the secondary regions replay the event to pick up the new files and drop the old ones.

Committing writes in the same order as in primary ensures that the secondaries won’t diverge from the primary regions data, but since the log replication is asynchronous, the data might still be stale in secondary regions. Since this feature works as a replication endpoint, the performance and latency characteristics is expected to be similar to inter-cluster replication.

Async WAL Replication is disabled by default. You can enable this feature by setting hbase.region.replica.replication.enabled to true. Asyn WAL Replication feature will add a new replication peer named region_replica_replication as a replication peer when you create a table with region replication > 1 for the first time. Once enabled, if you want to disable this feature, you need to do two actions: * Set configuration property hbase.region.replica.replication.enabled to false in hbase-site.xml (see Configuration section below) * Disable the replication peer named region_replica_replication in the cluster using hbase shell or ReplicationAdmin class:

        hbase> disable_peer 'region_replica_replication'

73.6. Store File TTL

In both of the write propagation approaches mentioned above, store files of the primary will be opened in secondaries independent of the primary region. So for files that the primary compacted away, the secondaries might still be referring to these files for reading. Both features are using HFileLinks to refer to files, but there is no protection (yet) for guaranteeing that the file will not be deleted prematurely. Thus, as a guard, you should set the configuration property hbase.master.hfilecleaner.ttl to a larger value, such as 1 hour to guarantee that you will not receive IOExceptions for requests going to replicas.

73.7. Region replication for META table’s region

Currently, Async WAL Replication is not done for the META table’s WAL. The meta table’s secondary replicas still refreshes themselves from the persistent store files. Hence the hbase.regionserver.meta.storefile.refresh.period needs to be set to a certain non-zero value for refreshing the meta store files. Note that this configuration is configured differently than hbase.regionserver.storefile.refresh.period.

73.8. Memory accounting

The secondary region replicas refer to the data files of the primary region replica, but they have their own memstores (in HBase-1.1+) and uses block cache as well. However, one distinction is that the secondary region replicas cannot flush the data when there is memory pressure for their memstores. They can only free up memstore memory when the primary region does a flush and this flush is replicated to the secondary. Since in a region server hosting primary replicas for some regions and secondaries for some others, the secondaries might cause extra flushes to the primary regions in the same host. In extreme situations, there can be no memory left for adding new writes coming from the primary via wal replication. For unblocking this situation (and since secondary cannot flush by itself), the secondary is allowed to do a “store file refresh” by doing a file system list operation to pick up new files from primary, and possibly dropping its memstore. This refresh will only be performed if the memstore size of the biggest secondary region replica is at least hbase.region.replica.storefile.refresh.memstore.multiplier (default 4) times bigger than the biggest memstore of a primary replica. One caveat is that if this is performed, the secondary can observe partial row updates across column families (since column families are flushed independently). The default should be good to not do this operation frequently. You can set this value to a large number to disable this feature if desired, but be warned that it might cause the replication to block forever.

73.9. Secondary replica failover

When a secondary region replica first comes online, or fails over, it may have served some edits from its memstore. Since the recovery is handled differently for secondary replicas, the secondary has to ensure that it does not go back in time before it starts serving requests after assignment. For doing that, the secondary waits until it observes a full flush cycle (start flush, commit flush) or a “region open event” replicated from the primary. Until this happens, the secondary region replica will reject all read requests by throwing an IOException with message “The region’s reads are disabled”. However, the other replicas will probably still be available to read, thus not causing any impact for the rpc with TIMELINE consistency. To facilitate faster recovery, the secondary region will trigger a flush request from the primary when it is opened. The configuration property hbase.region.replica.wait.for.primary.flush (enabled by default) can be used to disable this feature if needed.

73.10. Configuration properties

To use highly available reads, you should set the following properties in hbase-site.xml file. There is no specific configuration to enable or disable region replicas. Instead you can change the number of region replicas per table to increase or decrease at the table creation or with alter table. The following configuration is for using async wal replication and using meta replicas of 3.

73.10.1. Server side properties

      The period (in milliseconds) for refreshing the store files for the secondary regions. 0 means this feature is disabled. Secondary regions sees new files (from flushes and compactions) from primary once the secondary region refreshes the list of files in the region (there is no notification mechanism). But too frequent refreshes might cause extra Namenode pressure. If the files cannot be refreshed for longer than HFile TTL (hbase.master.hfilecleaner.ttl) the requests are rejected. Configuring HFile TTL to a larger value is also recommended with this setting.

      The period (in milliseconds) for refreshing the store files for the hbase:meta tables secondary regions. 0 means this feature is disabled. Secondary regions sees new files (from flushes and compactions) from primary once the secondary region refreshes the list of files in the region (there is no notification mechanism). But too frequent refreshes might cause extra Namenode pressure. If the files cannot be refreshed for longer than HFile TTL (hbase.master.hfilecleaner.ttl) the requests are rejected. Configuring HFile TTL to a larger value is also recommended with this setting. This should be a non-zero number if meta replicas are enabled (via hbase.meta.replica.count set to greater than 1).

      Whether asynchronous WAL replication to the secondary region replicas is enabled or not. If this is enabled, a replication peer named "region_replica_replication" will be created which will tail the logs and replicate the mutations to region replicas for tables that have region replication > 1. If this is enabled once, disabling this replication also      requires disabling the replication peer using shell or ReplicationAdmin java class. Replication to secondary region replicas works over standard inter-cluster replication. So replication, if disabled explicitly, also has to be enabled by setting "hbase.replication"· to true for this feature to work.
    If you set this to `false`, replicas do not receive memstore updates from
    the primary RegionServer. If you set this to `true`, you can still disable
    memstore replication on a per-table basis, by setting the table's
    `REGION_MEMSTORE_REPLICATION` configuration property to `false`. If
    memstore replication is disabled, the secondaries will only receive
    updates for events like flushes and bulkloads, and will not have access to
    data which the primary has not yet flushed. This preserves the guarantee
    of row-level consistency, even when the read requests `Consistency.TIMELINE`.

      The period (in milliseconds) to keep store files in the archive folder before deleting them from the file system.</description>

      Region replication count for the meta regions. Defaults to 1.

      The multiplier for a “store file refresh” operation for the secondary region replica. If a region server has memory pressure, the secondary region will refresh it’s store files if the memstore size of the biggest secondary replica is bigger this many times than the memstore size of the biggest primary replica. Set this to a very big value to disable this feature (not recommended).

      Whether to wait for observing a full flush cycle from the primary before start serving data in a secondary. Disabling this might cause the secondary region replicas to go back in time for reads between region movements.

One thing to keep in mind also is that, region replica placement policy is only enforced by the StochasticLoadBalancer which is the default balancer. If you are using a custom load balancer property in hbase-site.xml (hbase.master.loadbalancer.class) replicas of regions might end up being hosted in the same server.

73.10.2. Client side properties

Ensure to set the following for all clients (and servers) that will use region replicas.

      Whether to enable interruption of RPC threads at the client side. This is required for region replicas with fallback RPC’s to secondary regions.
    The timeout (in microseconds), before secondary fallback RPC’s are submitted for get requests with Consistency.TIMELINE to the secondary replicas of the regions. Defaults to 10ms. Setting this lower will increase the number of RPC’s, but will lower the p99 latencies.
      The timeout (in microseconds), before secondary fallback RPC’s are submitted for multi-get requests (Table.get(List<Get>)) with Consistency.TIMELINE to the secondary replicas of the regions. Defaults to 10ms. Setting this lower will increase the number of RPC’s, but will lower the p99 latencies.
    The timeout (in microseconds), before secondary fallback RPC’s are submitted for scan requests with Consistency.TIMELINE to the secondary replicas of the regions. Defaults to 1 sec. Setting this lower will increase the number of RPC’s, but will lower the p99 latencies.
      Whether to use meta table replicas or not. Default is false.

Note HBase-1.0.x users should use hbase.ipc.client.allowsInterrupt rather than hbase.ipc.client.specificThreadForWriting.

73.11. User Interface

In the masters user interface, the region replicas of a table are also shown together with the primary regions. You can notice that the replicas of a region will share the same start and end keys and the same region name prefix. The only difference would be the appended replica_id (which is encoded as hex), and the region encoded name will be different. You can also see the replica ids shown explicitly in the UI.

73.12. Creating a table with region replication

Region replication is a per-table property. All tables have REGION_REPLICATION = 1 by default, which means that there is only one replica per region. You can set and change the number of replicas per region of a table by supplying the REGION_REPLICATION property in the table descriptor.

73.12.1. Shell

create 't1', 'f1', {REGION_REPLICATION => 2}

describe 't1'
for i in 1..100
put 't1', "r#{i}", 'f1:c1', i
flush 't1'

73.12.2. Java

HTableDescriptor htd = new HTableDescriptor(TableName.valueOf(test_table));

You can also use setRegionReplication() and alter table to increase, decrease the region replication for a table.

73.13. Read API and Usage

73.13.1. Shell

You can do reads in shell using a the Consistency.TIMELINE semantics as follows

hbase(main):001:0> get 't1','r6', {CONSISTENCY => "TIMELINE"}

You can simulate a region server pausing or becoming unavailable and do a read from the secondary replica:

$ kill -STOP <pid or primary region server>

hbase(main):001:0> get 't1','r6', {CONSISTENCY => "TIMELINE"}

Using scans is also similar

hbase> scan 't1', {CONSISTENCY => 'TIMELINE'}

73.13.2. Java

You can set the consistency for Gets and Scans and do requests as follows.

Get get = new Get(row);
Result result = table.get(get);

You can also pass multiple gets:

Get get1 = new Get(row);
ArrayList<Get> gets = new ArrayList<Get>();
Result[] results = table.get(gets);

And Scans:

Scan scan = new Scan();
ResultScanner scanner = table.getScanner(scan);

You can inspect whether the results are coming from primary region or not by calling the Result.isStale() method:

Result result = table.get(get);
if (result.isStale()) {

73.14. Resources

  1. More information about the design and implementation can be found at the jira issue: HBASE-10070

  2. HBaseCon 2014 talk also contains some details and slides.

74. Storing Medium-sized Objects (MOB)

Data comes in many sizes, and saving all of your data in HBase, including binary data such as images and documents, is ideal. While HBase can technically handle binary objects with cells that are larger than 100 KB in size, HBase’s normal read and write paths are optimized for values smaller than 100KB in size. When HBase deals with large numbers of objects over this threshold, referred to here as medium objects, or MOBs, performance is degraded due to write amplification caused by splits and compactions. When using MOBs, ideally your objects will be between 100KB and 10MB. HBase FIX_VERSION_NUMBER adds support for better managing large numbers of MOBs while maintaining performance, consistency, and low operational overhead. MOB support is provided by the work done in HBASE-11339. To take advantage of MOB, you need to use HFile version 3. Optionally, configure the MOB file reader’s cache settings for each RegionServer (see Configuring the MOB Cache), then configure specific columns to hold MOB data. Client code does not need to change to take advantage of HBase MOB support. The feature is transparent to the client.

74.1. Configuring Columns for MOB

You can configure columns to support MOB during table creation or alteration, either in HBase Shell or via the Java API. The two relevant properties are the boolean IS_MOB and the MOB_THRESHOLD, which is the number of bytes at which an object is considered to be a MOB. Only IS_MOB is required. If you do not specify the MOB_THRESHOLD, the default threshold value of 100 KB is used.

Example 38. Configure a Column for MOB Using HBase Shell
hbase> create 't1', {NAME => 'f1', IS_MOB => true, MOB_THRESHOLD => 102400}
hbase> alter 't1', {NAME => 'f1', IS_MOB => true, MOB_THRESHOLD => 102400}
Example 39. Configure a Column for MOB Using the Java API
HColumnDescriptor hcd = new HColumnDescriptor(f);

74.2. Testing MOB

The utility org.apache.hadoop.hbase.IntegrationTestIngestMOB is provided to assist with testing the MOB feature. The utility is run as follows:

$ sudo -u hbase hbase org.apache.hadoop.hbase.IntegrationTestIngestMOB \
            -threshold 102400 \
            -minMobDataSize 512 \
            -maxMobDataSize 5120
  • threshold is the threshold at which cells are considered to be MOBs. The default is 1 kB, expressed in bytes.

  • minMobDataSize is the minimum value for the size of MOB data. The default is 512 B, expressed in bytes.

  • maxMobDataSize is the maximum value for the size of MOB data. The default is 5 kB, expressed in bytes.

74.3. Configuring the MOB Cache

Because there can be a large number of MOB files at any time, as compared to the number of HFiles, MOB files are not always kept open. The MOB file reader cache is a LRU cache which keeps the most recently used MOB files open. To configure the MOB file reader’s cache on each RegionServer, add the following properties to the RegionServer’s hbase-site.xml, customize the configuration to suit your environment, and restart or rolling restart the RegionServer.

Example 40. Example MOB Cache Configuration
      Number of opened file handlers to cache.
      A larger value will benefit reads by providing more file handlers per mob
      file cache and would reduce frequent file opening and closing.
      However, if this is set too high, this could lead to a "too many opened file handers"
      The default value is 1000.
      The amount of time in seconds after which an unused file is evicted from the
      MOB cache. The default value is 3600 seconds.
      A multiplier (between 0.0 and 1.0), which determines how many files remain cached
      after the threshold of files that remains cached after a cache eviction occurs
      which is triggered by reaching the `hbase.mob.file.cache.size` threshold.
      The default value is 0.5f, which means that half the files (the least-recently-used
      ones) are evicted.

74.4. MOB Optimization Tasks

74.4.1. Manually Compacting MOB Files

To manually compact MOB files, rather than waiting for the configuration to trigger compaction, use the compact_mob or major_compact_mob HBase shell commands. These commands require the first argument to be the table name, and take an optional column family as the second argument. If the column family is omitted, all MOB-enabled column families are compacted.

hbase> compact_mob 't1', 'c1'
hbase> compact_mob 't1'
hbase> major_compact_mob 't1', 'c1'
hbase> major_compact_mob 't1'

These commands are also available via Admin.compactMob and Admin.majorCompactMob methods.

74.4.2. MOB Sweeper

HBase MOB a MapReduce job called the Sweeper tool for optimization. The Sweeper tool coalesces small MOB files or MOB files with many deletions or updates. The Sweeper tool is not required if you use native MOB compaction, which does not rely on MapReduce.

To configure the Sweeper tool, set the following options:

      If there are too many cells deleted in a mob file, it's regarded
      as an invalid file and needs to be merged.
      If existingCellsSize/mobFileSize is less than ratio, it's regarded
      as an invalid file. The default value is 0.5f.
      If the size of a mob file is less than this value, it's regarded as a small
      file and needs to be merged. The default value is 128MB.
      The flush size for the memstore used by sweep job. Each sweep reducer owns such a memstore.
      The default value is 128MB.
      The period that ExpiredMobFileCleanerChore runs. The unit is second.
      The default value is one day.

Next, add the HBase install directory, `$HBASE_HOME`/*, and HBase library directory to yarn-site.xml Adjust this example to suit your environment.

    <description>Classpath for typical applications.</description>
        $HBASE_HOME/*, $HBASE_HOME/lib/*

Finally, run the sweeper tool for each column which is configured for MOB.

$ org.apache.hadoop.hbase.mob.compactions.Sweeper _tableName_ _familyName_

Apache HBase APIs

This chapter provides information about performing operations using HBase native APIs. This information is not exhaustive, and provides a quick reference in addition to the User API Reference. The examples here are not comprehensive or complete, and should be used for purposes of illustration only.

Apache HBase also works with multiple external APIs. See Apache HBase External APIs for more information.

75. Examples

Example 41. Create, modify and delete a Table Using Java
package com.example.hbase.admin;

package util;

import java.io.IOException;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.HBaseConfiguration;
import org.apache.hadoop.hbase.HColumnDescriptor;
import org.apache.hadoop.hbase.HConstants;
import org.apache.hadoop.hbase.HTableDescriptor;
import org.apache.hadoop.hbase.TableName;
import org.apache.hadoop.hbase.client.Admin;
import org.apache.hadoop.hbase.client.Connection;
import org.apache.hadoop.hbase.client.ConnectionFactory;
import org.apache.hadoop.hbase.io.compress.Compression.Algorithm;

public class Example {

  private static final String TABLE_NAME = "MY_TABLE_NAME_TOO";
  private static final String CF_DEFAULT = "DEFAULT_COLUMN_FAMILY";

  public static void createOrOverwrite(Admin admin, HTableDescriptor table) throws IOException {
    if (admin.tableExists(table.getTableName())) {

  public static void createSchemaTables(Configuration config) throws IOException {
    try (Connection connection = ConnectionFactory.createConnection(config);
         Admin admin = connection.getAdmin()) {

      HTableDescriptor table = new HTableDescriptor(TableName.valueOf(TABLE_NAME));
      table.addFamily(new HColumnDescriptor(CF_DEFAULT).setCompressionType(Algorithm.SNAPPY));

      System.out.print("Creating table. ");
      createOrOverwrite(admin, table);
      System.out.println(" Done.");

  public static void modifySchema (Configuration config) throws IOException {
    try (Connection connection = ConnectionFactory.createConnection(config);
         Admin admin = connection.getAdmin()) {

      TableName tableName = TableName.valueOf(TABLE_NAME);
      if (admin.tableExists(tableName)) {
        System.out.println("Table does not exist.");

      HTableDescriptor table = new HTableDescriptor(tableName);

      // Update existing table
      HColumnDescriptor newColumn = new HColumnDescriptor("NEWCF");
      admin.addColumn(tableName, newColumn);

      // Update existing column family
      HColumnDescriptor existingColumn = new HColumnDescriptor(CF_DEFAULT);
      admin.modifyTable(tableName, table);

      // Disable an existing table

      // Delete an existing column family
      admin.deleteColumn(tableName, CF_DEFAULT.getBytes("UTF-8"));

      // Delete a table (Need to be disabled first)

  public static void main(String... args) throws IOException {
    Configuration config = HBaseConfiguration.create();

    //Add any necessary configuration files (hbase-site.xml, core-site.xml)
    config.addResource(new Path(System.getenv("HBASE_CONF_DIR"), "hbase-site.xml"));
    config.addResource(new Path(System.getenv("HADOOP_CONF_DIR"), "core-site.xml"));

Apache HBase External APIs

This chapter will cover access to Apache HBase either through non-Java languages and through custom protocols. For information on using the native HBase APIs, refer to User API Reference and the HBase APIs chapter.

76. REST

Representational State Transfer (REST) was introduced in 2000 in the doctoral dissertation of Roy Fielding, one of the principal authors of the HTTP specification.

REST itself is out of the scope of this documentation, but in general, REST allows client-server interactions via an API that is tied to the URL itself. This section discusses how to configure and run the REST server included with HBase, which exposes HBase tables, rows, cells, and metadata as URL specified resources. There is also a nice series of blogs on How-to: Use the Apache HBase REST Interface by Jesse Anderson.

76.1. Starting and Stopping the REST Server

The included REST server can run as a daemon which starts an embedded Jetty servlet container and deploys the servlet into it. Use one of the following commands to start the REST server in the foreground or background. The port is optional, and defaults to 8080.

# Foreground
$ bin/hbase rest start -p <port>

# Background, logging to a file in $HBASE_LOGS_DIR
$ bin/hbase-daemon.sh start rest -p <port>

To stop the REST server, use Ctrl-C if you were running it in the foreground, or the following command if you were running it in the background.

$ bin/hbase-daemon.sh stop rest

76.2. Configuring the REST Server and Client

For information about configuring the REST server and client for SSL, as well as doAs impersonation for the REST server, see Configure the Thrift Gateway to Authenticate on Behalf of the Client and other portions of the Securing Apache HBase chapter.

76.3. Using REST Endpoints

The following examples use the placeholder server http://example.com:8000, and the following commands can all be run using curl or wget commands. You can request plain text (the default), XML , or JSON output by adding no header for plain text, or the header "Accept: text/xml" for XML or "Accept: application/json" for JSON.

Unless specified, use GET requests for queries, PUT or POST requests for creation or mutation, and DELETE for deletion.

76.3.1. Cluster Information

HBase Version
Cluster Status
Table List

76.3.2. Table Information

Table Schema (GET)

To retrieve the table schema, use a GET request with the /schema endpoint:

Table Creation

To create a table, use a PUT request with the /schema endpoint:

Table Schema Update

To update a table, use a POST request with the /schema endpoint:

Table Deletion

To delete a table, use a DELETE request with the /schema endpoint:

Table Regions

76.3.3. Gets

GET a Single Cell Value

To get a single cell value, use a URL scheme like the following:


The column qualifier and timestamp are optional. Without them, the whole row will be returned, or the newest version will be returned.

Multiple Single Values (Multi-Get)

To get multiple single values, specify multiple column:qualifier tuples and/or a start-timestamp and end-timestamp. You can also limit the number of versions.

Globbing Rows

To scan a series of rows, you can use a * glob character on the <row> value to glob together multiple rows.


76.3.4. Puts

For Puts, PUT and POST are equivalent.

Put a Single Value

The column qualifier and the timestamp are optional.

Put Multiple Values

To put multiple values, use a false row key. Row, column, and timestamp values in the supplied cells override the specifications on the path, allowing you to post multiple values to a table in batch. The HTTP response code indicates the status of the put. Set the Content-Type to text/xml for XML encoding or to application/x-protobuf for protobufs encoding. Supply the commit data in the PUT or POST body, using the REST XML Schema and REST Protobufs Schema as guidelines.

76.3.5. Scans

PUT and POST are equivalent for scans.

Scanner Creation

To create a scanner, use the /scanner endpoint. The HTTP response code indicates success (201) or failure (anything else), and on successful scanner creation, the URI is returned which should be used to address the scanner.

Scanner Get Next

To get the next batch of cells found by the scanner, use the /scanner/<scanner-id>' endpoint, using the URI returned by the scanner creation endpoint. If the scanner is exhausted, HTTP status `204 is returned.

Scanner Deletion

To delete resources associated with a scanner, send a HTTP DELETE request to the /scanner/<scanner-id> endpoint.


76.4. REST XML Schema

<schema xmlns="http://www.w3.org/2001/XMLSchema" xmlns:tns="RESTSchema">

  <element name="Version" type="tns:Version"></element>

  <complexType name="Version">
    <attribute name="REST" type="string"></attribute>
    <attribute name="JVM" type="string"></attribute>
    <attribute name="OS" type="string"></attribute>
    <attribute name="Server" type="string"></attribute>
    <attribute name="Jersey" type="string"></attribute>

  <element name="TableList" type="tns:TableList"></element>

  <complexType name="TableList">
      <element name="table" type="tns:Table" maxOccurs="unbounded" minOccurs="1"></element>

  <complexType name="Table">
      <element name="name" type="string"></element>

  <element name="TableInfo" type="tns:TableInfo"></element>

  <complexType name="TableInfo">
      <element name="region" type="tns:TableRegion" maxOccurs="unbounded" minOccurs="1"></element>
    <attribute name="name" type="string"></attribute>

  <complexType name="TableRegion">
    <attribute name="name" type="string"></attribute>
    <attribute name="id" type="int"></attribute>
    <attribute name="startKey" type="base64Binary"></attribute>
    <attribute name="endKey" type="base64Binary"></attribute>
    <attribute name="location" type="string"></attribute>

  <element name="TableSchema" type="tns:TableSchema"></element>

  <complexType name="TableSchema">
      <element name="column" type="tns:ColumnSchema" maxOccurs="unbounded" minOccurs="1"></element>
    <attribute name="name" type="string"></attribute>

  <complexType name="ColumnSchema">
    <attribute name="name" type="string"></attribute>

  <element name="CellSet" type="tns:CellSet"></element>

  <complexType name="CellSet">
      <element name="row" type="tns:Row" maxOccurs="unbounded" minOccurs="1"></element>

  <element name="Row" type="tns:Row"></element>

  <complexType name="Row">
      <element name="key" type="base64Binary"></element>
      <element name="cell" type="tns:Cell" maxOccurs="unbounded" minOccurs="1"></element>

  <element name="Cell" type="tns:Cell"></element>

  <complexType name="Cell">
      <element name="value" maxOccurs="1" minOccurs="1">
        <simpleType><restriction base="base64Binary">
    <attribute name="column" type="base64Binary" />
    <attribute name="timestamp" type="int" />

  <element name="Scanner" type="tns:Scanner"></element>

  <complexType name="Scanner">
      <element name="column" type="base64Binary" minOccurs="0" maxOccurs="unbounded"></element>
      <element name="filter" type="string" minOccurs="0" maxOccurs="1"></element>
    <attribute name="startRow" type="base64Binary"></attribute>
    <attribute name="endRow" type="base64Binary"></attribute>
    <attribute name="batch" type="int"></attribute>
    <attribute name="startTime" type="int"></attribute>
    <attribute name="endTime" type="int"></attribute>

  <element name="StorageClusterVersion" type="tns:StorageClusterVersion" />

  <complexType name="StorageClusterVersion">
    <attribute name="version" type="string"></attribute>

  <element name="StorageClusterStatus"

  <complexType name="StorageClusterStatus">
      <element name="liveNode" type="tns:Node"
        maxOccurs="unbounded" minOccurs="0">
      <element name="deadNode" type="string" maxOccurs="unbounded"
    <attribute name="regions" type="int"></attribute>
    <attribute name="requests" type="int"></attribute>
    <attribute name="averageLoad" type="float"></attribute>

  <complexType name="Node">
      <element name="region" type="tns:Region"
          maxOccurs="unbounded" minOccurs="0">
    <attribute name="name" type="string"></attribute>
    <attribute name="startCode" type="int"></attribute>
    <attribute name="requests" type="int"></attribute>
    <attribute name="heapSizeMB" type="int"></attribute>
    <attribute name="maxHeapSizeMB" type="int"></attribute>

  <complexType name="Region">
    <attribute name="name" type="base64Binary"></attribute>
    <attribute name="stores" type="int"></attribute>
    <attribute name="storefiles" type="int"></attribute>
    <attribute name="storefileSizeMB" type="int"></attribute>
    <attribute name="memstoreSizeMB" type="int"></attribute>
    <attribute name="storefileIndexSizeMB" type="int"></attribute>


76.5. REST Protobufs Schema

message Version {
  optional string restVersion = 1;
  optional string jvmVersion = 2;
  optional string osVersion = 3;
  optional string serverVersion = 4;
  optional string jerseyVersion = 5;

message StorageClusterStatus {
  message Region {
    required bytes name = 1;
    optional int32 stores = 2;
    optional int32 storefiles = 3;
    optional int32 storefileSizeMB = 4;
    optional int32 memstoreSizeMB = 5;
    optional int32 storefileIndexSizeMB = 6;
  message Node {
    required string name = 1;    // name:port
    optional int64 startCode = 2;
    optional int32 requests = 3;
    optional int32 heapSizeMB = 4;
    optional int32 maxHeapSizeMB = 5;
    repeated Region regions = 6;
  // node status
  repeated Node liveNodes = 1;
  repeated string deadNodes = 2;
  // summary statistics
  optional int32 regions = 3;
  optional int32 requests = 4;
  optional double averageLoad = 5;

message TableList {
  repeated string name = 1;

message TableInfo {
  required string name = 1;
  message Region {
    required string name = 1;
    optional bytes startKey = 2;
    optional bytes endKey = 3;
    optional int64 id = 4;
    optional string location = 5;
  repeated Region regions = 2;

message TableSchema {
  optional string name = 1;
  message Attribute {
    required string name = 1;
    required string value = 2;
  repeated Attribute attrs = 2;
  repeated ColumnSchema columns = 3;
  // optional helpful encodings of commonly used attributes
  optional bool inMemory = 4;
  optional bool readOnly = 5;

message ColumnSchema {
  optional string name = 1;
  message Attribute {
    required string name = 1;
    required string value = 2;
  repeated Attribute attrs = 2;
  // optional helpful encodings of commonly used attributes
  optional int32 ttl = 3;
  optional int32 maxVersions = 4;
  optional string compression = 5;

message Cell {
  optional bytes row = 1;       // unused if Cell is in a CellSet
  optional bytes column = 2;
  optional int64 timestamp = 3;
  optional bytes data = 4;

message CellSet {
  message Row {
    required bytes key = 1;
    repeated Cell values = 2;
  repeated Row rows = 1;

message Scanner {
  optional bytes startRow = 1;
  optional bytes endRow = 2;
  repeated bytes columns = 3;
  optional int32 batch = 4;
  optional int64 startTime = 5;
  optional int64 endTime = 6;

77. Thrift

Documentation about Thrift has moved to Thrift API and Filter Language.

78. C/C++ Apache HBase Client

FB’s Chip Turner wrote a pure C/C++ client. Check it out.

79. Using Java Data Objects (JDO) with HBase

Java Data Objects (JDO) is a standard way to access persistent data in databases, using plain old Java objects (POJO) to represent persistent data.


This code example has the following dependencies:

  1. HBase 0.90.x or newer

  2. commons-beanutils.jar (http://commons.apache.org/)

  3. commons-pool-1.5.5.jar (http://commons.apache.org/)

  4. transactional-tableindexed for HBase 0.90 (https://github.com/hbase-trx/hbase-transactional-tableindexed)

Download hbase-jdo

Download the code from http://code.google.com/p/hbase-jdo/.

Example 42. JDO Example

This example uses JDO to create a table and an index, insert a row into a table, get a row, get a column value, perform a query, and do some additional HBase operations.

package com.apache.hadoop.hbase.client.jdo.examples;

import java.io.File;
import java.io.FileInputStream;
import java.io.InputStream;
import java.util.Hashtable;

import org.apache.hadoop.fs.Path;
import org.apache.hadoop.hbase.client.tableindexed.IndexedTable;

import com.apache.hadoop.hbase.client.jdo.AbstractHBaseDBO;
import com.apache.hadoop.hbase.client.jdo.HBaseBigFile;
import com.apache.hadoop.hbase.client.jdo.HBaseDBOImpl;
import com.apache.hadoop.hbase.client.jdo.query.DeleteQuery;
import com.apache.hadoop.hbase.client.jdo.query.HBaseOrder;
import com.apache.hadoop.hbase.client.jdo.query.HBaseParam;
import com.apache.hadoop.hbase.client.jdo.query.InsertQuery;
import com.apache.hadoop.hbase.client.jdo.query.QSearch;
import com.apache.hadoop.hbase.client.jdo.query.SelectQuery;
import com.apache.hadoop.hbase.client.jdo.query.UpdateQuery;

 * Hbase JDO Example.
 * dependency library.
 * - commons-beanutils.jar
 * - commons-pool-1.5.5.jar
 * - hbase0.90.0-transactionl.jar
 * you can expand Delete,Select,Update,Insert Query classes.
public class HBaseExample {
  public static void main(String[] args) throws Exception {
    AbstractHBaseDBO dbo = new HBaseDBOImpl();

    //*drop if table is already exist.*

    //*create table*

    //create index.
    String[] cols={"id","name"};

    InsertQuery insert = dbo.createInsertQuery("user");
    UserBean bean = new UserBean();

    //select 1 row
    SelectQuery select = dbo.createSelectQuery("user");
    UserBean resultBean = (UserBean)select.select(bean.getRow(),UserBean.class);

    // select column value.
    String value = (String)select.selectColumn(bean.getRow(),"account","id",String.class);

    // search with option (QSearch has EQUAL, NOT_EQUAL, LIKE)
    // select id,password,name,email from account where id='ncanis' limit startRow,20
    HBaseParam param = new HBaseParam();
    select.search("account", param, UserBean.class);

    // search column value is existing.
    boolean isExist = select.existColumnValue("account","id","ncanis".getBytes());

    // update password.
    UpdateQuery update = dbo.createUpdateQuery("user");
    Hashtable<String, byte[]> colsTable = new Hashtable<String, byte[]>();

    DeleteQuery delete = dbo.createDeleteQuery("user");

    // etc

    // HTable pool with apache commons pool
    // borrow and release. HBasePoolManager(maxActive, minIdle etc..)
    IndexedTable table = dbo.getPool().borrow("user");

    // upload bigFile by hadoop directly.
    HBaseBigFile bigFile = new HBaseBigFile();
    File file = new File("doc/movie.avi");
    FileInputStream fis = new FileInputStream(file);
    Path rootPath = new Path("/files/");
    String filename = "movie.avi";

    // receive file stream from hadoop.
    Path p = new Path(rootPath,filename);
    InputStream is = bigFile.path2Stream(p,4096);


80. Scala

80.1. Setting the Classpath

To use Scala with HBase, your CLASSPATH must include HBase’s classpath as well as the Scala JARs required by your code. First, use the following command on a server running the HBase RegionServer process, to get HBase’s classpath.

$ ps aux |grep regionserver| awk -F 'java.library.path=' {'print $2'} | awk {'print $1'}


Set the $CLASSPATH environment variable to include the path you found in the previous step, plus the path of scala-library.jar and each additional Scala-related JAR needed for your project.

$ export CLASSPATH=$CLASSPATH:/usr/lib/hadoop/lib/native:/usr/lib/hbase/lib/native/Linux-amd64-64:/path/to/scala-library.jar

80.2. Scala SBT File

Your build.sbt file needs the following resolvers and libraryDependencies to work with HBase.

resolvers += "Apache HBase" at "https://repository.apache.org/content/repositories/releases"

resolvers += "Thrift" at "http://people.apache.org/~rawson/repo/"

libraryDependencies ++= Seq(
    "org.apache.hadoop" % "hadoop-core" % "0.20.2",
    "org.apache.hbase" % "hbase" % "0.90.4"

80.3. Example Scala Code

This example lists HBase tables, creates a new table, and adds a row to it.

import org.apache.hadoop.hbase.HBaseConfiguration
import org.apache.hadoop.hbase.client.{Connection,ConnectionFactory,HBaseAdmin,HTable,Put,Get}
import org.apache.hadoop.hbase.util.Bytes

val conf = new HBaseConfiguration()
val connection = ConnectionFactory.createConnection(conf);
val admin = connection.getAdmin();

// list the tables
val listtables=admin.listTables()

// let's insert some data in 'mytable' and get the row

val table = new HTable(conf, "mytable")

val theput= new Put(Bytes.toBytes("rowkey1"))


val theget= new Get(Bytes.toBytes("rowkey1"))
val result=table.get(theget)
val value=result.value()

81. Jython

81.1. Setting the Classpath

To use Jython with HBase, your CLASSPATH must include HBase’s classpath as well as the Jython JARs required by your code. First, use the following command on a server running the HBase RegionServer process, to get HBase’s classpath.

$ ps aux |grep regionserver| awk -F 'java.library.path=' {'print $2'} | awk {'print $1'}


Set the $CLASSPATH environment variable to include the path you found in the previous step, plus the path to jython.jar and each additional Jython-related JAR needed for your project.

$ export CLASSPATH=$CLASSPATH:/usr/lib/hadoop/lib/native:/usr/lib/hbase/lib/native/Linux-amd64-64:/path/to/jython.jar

Start a Jython shell with HBase and Hadoop JARs in the classpath: $ bin/hbase org.python.util.jython

81.2. Jython Code Examples

Example 43. Table Creation, Population, Get, and Delete with Jython

The following Jython code example creates a table, populates it with data, fetches the data, and deletes the table.

import java.lang
from org.apache.hadoop.hbase import HBaseConfiguration, HTableDescriptor, HColumnDescriptor, HConstants, TableName
from org.apache.hadoop.hbase.client import HBaseAdmin, HTable, Get
from org.apache.hadoop.hbase.io import Cell, RowResult

# First get a conf object.  This will read in the configuration
# that is out in your hbase-*.xml files such as location of the
# hbase master node.
conf = HBaseConfiguration()

# Create a table named 'test' that has two column families,
# one named 'content, and the other 'anchor'.  The colons
# are required for column family names.
tablename = TableName.valueOf("test")

desc = HTableDescriptor(tablename)
admin = HBaseAdmin(conf)

# Drop and recreate if it exists
if admin.tableExists(tablename):

tables = admin.listTables()
table = HTable(conf, tablename)

# Add content to 'column:' on a row named 'row_x'
row = 'row_x'
update = Get(row)
update.put('content:', 'some content')

# Now fetch the content just added, returns a byte[]
data_row = table.get(row, "content:")
data = java.lang.String(data_row.value, "UTF8")

print "The fetched row contains the value '%s'" % data

# Delete the table.
Example 44. Table Scan Using Jython

This example scans a table and returns the results that match a given family qualifier.

# Print all rows that are members of a particular column family
# by passing a regex for family qualifier

import java.lang

from org.apache.hadoop.hbase import HBaseConfiguration
from org.apache.hadoop.hbase.client import HTable

conf = HBaseConfiguration()

table = HTable(conf, "wiki")
col = "title:.*$"

scanner = table.getScanner([col], "")
while 1:
    result = scanner.next()
    if not result:
    print java.lang.String(result.row), java.lang.String(result.get('title:').value)

Thrift API and Filter Language

Apache Thrift is a cross-platform, cross-language development framework. HBase includes a Thrift API and filter language. The Thrift API relies on client and server processes.

You can configure Thrift for secure authentication at the server and client side, by following the procedures in Client-side Configuration for Secure Operation - Thrift Gateway and Configure the Thrift Gateway to Authenticate on Behalf of the Client.

The rest of this chapter discusses the filter language provided by the Thrift API.

82. Filter Language

Thrift Filter Language was introduced in HBase 0.92. It allows you to perform server-side filtering when accessing HBase over Thrift or in the HBase shell. You can find out more about shell integration by using the scan help command in the shell.

You specify a filter as a string, which is parsed on the server to construct the filter.

82.1. General Filter String Syntax

A simple filter expression is expressed as a string:

“FilterName (argument, argument,... , argument)”

Keep the following syntax guidelines in mind.

  • Specify the name of the filter followed by the comma-separated argument list in parentheses.

  • If the argument represents a string, it should be enclosed in single quotes (').

  • Arguments which represent a boolean, an integer, or a comparison operator (such as <, >, or !=), should not be enclosed in quotes

  • The filter name must be a single word. All ASCII characters are allowed except for whitespace, single quotes and parentheses.

  • The filter’s arguments can contain any ASCII character. If single quotes are present in the argument, they must be escaped by an additional preceding single quote.

82.2. Compound Filters and Operators

Binary Operators

If the AND operator is used, the key-value must satisfy both filters.


If the OR operator is used, the key-value must satisfy at least one of the filters.

Unary Operators

For a particular row, if any of the key-values fail the filter condition, the entire row is skipped.


For a particular row, key-values will be emitted until a key-value is reached that fails the filter condition.

Example 45. Compound Operators

You can combine multiple operators to create a hierarchy of filters, such as the following example:

(Filter1 AND Filter2) OR (Filter3 AND Filter4)

82.3. Order of Evaluation

  1. Parentheses have the highest precedence.

  2. The unary operators SKIP and WHILE are next, and have the same precedence.

  3. The binary operators follow. AND has highest precedence, followed by OR.

Example 46. Precedence Example
Filter1 AND Filter2 OR Filter
is evaluated as
(Filter1 AND Filter2) OR Filter3
Filter1 AND SKIP Filter2 OR Filter3
is evaluated as
(Filter1 AND (SKIP Filter2)) OR Filter3

You can use parentheses to explicitly control the order of evaluation.

82.4. Compare Operator

The following compare operators are provided:

  1. LESS (<)

  2. LESS_OR_EQUAL (⇐)

  3. EQUAL (=)

  4. NOT_EQUAL (!=)


  6. GREATER (>)

  7. NO_OP (no operation)

The client should use the symbols (<, ⇐, =, !=, >, >=) to express compare operators.

82.5. Comparator

A comparator can be any of the following:

  1. BinaryComparator - This lexicographically compares against the specified byte array using Bytes.compareTo(byte[], byte[])

  2. BinaryPrefixComparator - This lexicographically compares against a specified byte array. It only compares up to the length of this byte array.

  3. RegexStringComparator - This compares against the specified byte array using the given regular expression. Only EQUAL and NOT_EQUAL comparisons are valid with this comparator

  4. SubStringComparator - This tests if the given substring appears in a specified byte array. The comparison is case insensitive. Only EQUAL and NOT_EQUAL comparisons are valid with this comparator

The general syntax of a comparator is: ComparatorType:ComparatorValue

The ComparatorType for the various comparators is as follows:

  1. BinaryComparator - binary

  2. BinaryPrefixComparator - binaryprefix

  3. RegexStringComparator - regexstring

  4. SubStringComparator - substring

The ComparatorValue can be any value.

Example ComparatorValues
  1. binary:abc will match everything that is lexicographically greater than "abc"

  2. binaryprefix:abc will match everything whose first 3 characters are lexicographically equal to "abc"

  3. regexstring:ab*yz will match everything that doesn’t begin with "ab" and ends with "yz"

  4. substring:abc123 will match everything that begins with the substring "abc123"

82.6. Example PHP Client Program that uses the Filter Language

  $_SERVER['PHP_ROOT'] = realpath(dirname(__FILE__).'/..');
  require_once $_SERVER['PHP_ROOT'].'/flib/__flib.php';
  $hbase = new HBase('<server_name_running_thrift_server>', <port on which thrift server is running>);
  $client = $hbase->getClient();
  $result = $client->scannerOpenWithFilterString('table_name', "(PrefixFilter ('row2') AND (QualifierFilter (>=, 'binary:xyz'))) AND (TimestampsFilter ( 123, 456))");
  $to_print = $client->scannerGetList($result,1);
  while ($to_print) {
    $to_print = $client->scannerGetList($result,1);

82.7. Example Filter Strings

  • "PrefixFilter ('Row') AND PageFilter (1) AND FirstKeyOnlyFilter ()" will return all key-value pairs that match the following conditions:

    1. The row containing the key-value should have prefix Row

    2. The key-value must be located in the first row of the table

    3. The key-value pair must be the first key-value in the row

  • "(RowFilter (=, 'binary:Row 1') AND TimeStampsFilter (74689, 89734)) OR ColumnRangeFilter ('abc', true, 'xyz', false))" will return all key-value pairs that match both the following conditions:

    • The key-value is in a row having row key Row 1

    • The key-value must have a timestamp of either 74689 or 89734.

    • Or it must match the following condition:

      • The key-value pair must be in a column that is lexicographically >= abc and < xyz 

  • "SKIP ValueFilter (0)" will skip the entire row if any of the values in the row is not 0

82.8. Individual Filter Syntax


This filter doesn’t take any arguments. It returns only the key component of each key-value.


This filter doesn’t take any arguments. It returns only the first key-value from each row.


This filter takes one argument – a prefix of a row key. It returns only those key-values present in a row that starts with the specified row prefix


This filter takes one argument – a column prefix. It returns only those key-values present in a column that starts with the specified column prefix. The column prefix must be of the form: “qualifier”.


This filter takes a list of column prefixes. It returns key-values that are present in a column that starts with any of the specified column prefixes. Each of the column prefixes must be of the form: “qualifier”.


This filter takes one argument – a limit. It returns the first limit number of columns in the table.


This filter takes one argument – a page size. It returns page size number of rows from the table.


This filter takes two arguments – a limit and offset. It returns limit number of columns after offset number of columns. It does this for all the rows.


This filter takes one argument – a row key on which to stop scanning. It returns all key-values present in rows up to and including the specified row.


This filter takes a list of timestamps. It returns those key-values whose timestamps matches any of the specified timestamps.


This filter takes a compare operator and a comparator. It compares each row key with the comparator using the compare operator and if the comparison returns true, it returns all the key-values in that row.

Family Filter

This filter takes a compare operator and a comparator. It compares each column family name with the comparator using the compare operator and if the comparison returns true, it returns all the Cells in that column family.


This filter takes a compare operator and a comparator. It compares each qualifier name with the comparator using the compare operator and if the comparison returns true, it returns all the key-values in that column.


This filter takes a compare operator and a comparator. It compares each value with the comparator using the compare operator and if the comparison returns true, it returns that key-value.


This filter takes two arguments – a family and a qualifier. It tries to locate this column in each row and returns all key-values in that row that have the same timestamp. If the row doesn’t contain the specified column – none of the key-values in that row will be returned.


This filter takes a column family, a qualifier, a compare operator and a comparator. If the specified column is not found – all the columns of that row will be emitted. If the column is found and the comparison with the comparator returns true, all the columns of the row will be emitted. If the condition fails, the row will not be emitted.


This filter takes the same arguments and behaves same as SingleColumnValueFilter – however, if the column is found and the condition passes, all the columns of the row will be emitted except for the tested column value.


This filter is used for selecting only those keys with columns that are between minColumn and maxColumn. It also takes two boolean variables to indicate whether to include the minColumn and maxColumn or not.

Apache HBase Coprocessors

HBase Coprocessors are modeled after Google BigTable’s coprocessor implementation (http://research.google.com/people/jeff/SOCC2010-keynote-slides.pdf pages 41-42.).

The coprocessor framework provides mechanisms for running your custom code directly on the RegionServers managing your data. Efforts are ongoing to bridge gaps between HBase’s implementation and BigTable’s architecture. For more information see HBASE-4047.

The information in this chapter is primarily sourced and heavily reused from the following resources:

  1. Mingjie Lai’s blog post Coprocessor Introduction.

  2. Gaurav Bhardwaj’s blog post The How To Of HBase Coprocessors.

Use Coprocessors At Your Own Risk

Coprocessors are an advanced feature of HBase and are intended to be used by system developers only. Because coprocessor code runs directly on the RegionServer and has direct access to your data, they introduce the risk of data corruption, man-in-the-middle attacks, or other malicious data access. Currently, there is no mechanism to prevent data corruption by coprocessors, though work is underway on HBASE-4047.

+ In addition, there is no resource isolation, so a well-intentioned but misbehaving coprocessor can severely degrade cluster performance and stability.

83. Coprocessor Overview

In HBase, you fetch data using a Get or Scan, whereas in an RDBMS you use a SQL query. In order to fetch only the relevant data, you filter it using a HBase Filter , whereas in an RDBMS you use a WHERE predicate.

After fetching the data, you perform computations on it. This paradigm works well for "small data" with a few thousand rows and several columns. However, when you scale to billions of rows and millions of columns, moving large amounts of data across your network will create bottlenecks at the network layer, and the client needs to be powerful enough and have enough memory to handle the large amounts of data and the computations. In addition, the client code can grow large and complex.

In this scenario, coprocessors might make sense. You can put the business computation code into a coprocessor which runs on the RegionServer, in the same location as the data, and returns the result to the client.

This is only one scenario where using coprocessors can provide benefit. Following are some analogies which may help to explain some of the benefits of coprocessors.

83.1. Coprocessor Analogies

Triggers and Stored Procedure

An Observer coprocessor is similar to a trigger in a RDBMS in that it executes your code either before or after a specific event (such as a Get or Put) occurs. An endpoint coprocessor is similar to a stored procedure in a RDBMS because it allows you to perform custom computations on the data on the RegionServer itself, rather than on the client.


MapReduce operates on the principle of moving the computation to the location of the data. Coprocessors operate on the same principal.


If you are familiar with Aspect Oriented Programming (AOP), you can think of a coprocessor as applying advice by intercepting a request and then running some custom code, before passing the request on to its final destination (or even changing the destination).

83.2. Coprocessor Implementation Overview

  1. Either your class should extend one of the Coprocessor classes, such as BaseRegionObserver, or it should implement the Coprocessor or CoprocessorService interface.

  2. Load the coprocessor, either statically (from the configuration) or dynamically, using HBase Shell. For more details see Loading Coprocessors.

  3. Call the coprocessor from your client-side code. HBase handles the coprocessor trapsparently.

The framework API is provided in the coprocessor package.

84. Types of Coprocessors

84.1. Observer Coprocessors

Observer coprocessors are triggered either before or after a specific event occurs. Observers that happen before an event use methods that start with a pre prefix, such as prePut. Observers that happen just after an event override methods that start with a post prefix, such as postPut.

84.1.1. Use Cases for Observer Coprocessors


Before performing a Get or <