001/*
002 * Licensed to the Apache Software Foundation (ASF) under one
003 * or more contributor license agreements.  See the NOTICE file
004 * distributed with this work for additional information
005 * regarding copyright ownership.  The ASF licenses this file
006 * to you under the Apache License, Version 2.0 (the
007 * "License"); you may not use this file except in compliance
008 * with the License.  You may obtain a copy of the License at
009 *
010 *     http://www.apache.org/licenses/LICENSE-2.0
011 *
012 * Unless required by applicable law or agreed to in writing, software
013 * distributed under the License is distributed on an "AS IS" BASIS,
014 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
015 * See the License for the specific language governing permissions and
016 * limitations under the License.
017 */
018package org.apache.hadoop.hbase.client;
019
020import java.io.Closeable;
021import java.io.IOException;
022import java.util.List;
023import java.util.Map;
024import java.util.stream.Collectors;
025import org.apache.hadoop.hbase.Abortable;
026import org.apache.hadoop.hbase.HBaseInterfaceAudience;
027import org.apache.hadoop.hbase.ServerName;
028import org.apache.hadoop.hbase.master.RegionState;
029import org.apache.yetus.audience.InterfaceAudience;
030
031import org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil;
032import org.apache.hadoop.hbase.shaded.protobuf.generated.HBaseProtos;
033
034/**
035 * Hbck fixup tool APIs. Obtain an instance from {@link ClusterConnection#getHbck()} and call
036 * {@link #close()} when done.
037 * <p>
038 * WARNING: the below methods can damage the cluster. It may leave the cluster in an indeterminate
039 * state, e.g. region not assigned, or some hdfs files left behind. After running any of the below,
040 * operators may have to do some clean up on hdfs or schedule some assign procedures to get regions
041 * back online. DO AT YOUR OWN RISK. For experienced users only.
042 * @see ConnectionFactory
043 * @see ClusterConnection
044 * @since 2.0.2, 2.1.1
045 */
046@InterfaceAudience.LimitedPrivate(HBaseInterfaceAudience.HBCK)
047public interface Hbck extends Abortable, Closeable {
048  /**
049   * Update table state in Meta only. No procedures are submitted to open/assign or close/unassign
050   * regions of the table.
051   * @param state table state
052   * @return previous state of the table in Meta
053   */
054  TableState setTableStateInMeta(TableState state) throws IOException;
055
056  /**
057   * Update region state in Meta only. No procedures are submitted to manipulate the given region or
058   * any other region from same table.
059   * @param nameOrEncodedName2State list of all region states to be updated in meta
060   * @return previous state of the region in Meta
061   */
062  Map<String, RegionState.State>
063    setRegionStateInMeta(Map<String, RegionState.State> nameOrEncodedName2State) throws IOException;
064
065  /**
066   * Like {@link Admin#assign(byte[])} but 'raw' in that it can do more than one Region at a time --
067   * good if many Regions to online -- and it will schedule the assigns even in the case where
068   * Master is initializing (as long as the ProcedureExecutor is up). Does NOT call Coprocessor
069   * hooks.
070   * @param override           You need to add the override for case where a region has previously
071   *                           been bypassed. When a Procedure has been bypassed, a Procedure will
072   *                           have completed but no other Procedure will be able to make progress
073   *                           on the target entity (intentionally). This override flag will
074   *                           override this fencing mechanism.
075   * @param encodedRegionNames Region encoded names; e.g. 1588230740 is the hard-coded encoding for
076   *                           hbase:meta region and de00010733901a05f5a2a3a382e27dd4 is an example
077   *                           of what a random user-space encoded Region name looks like.
078   */
079  List<Long> assigns(List<String> encodedRegionNames, boolean override) throws IOException;
080
081  default List<Long> assigns(List<String> encodedRegionNames) throws IOException {
082    return assigns(encodedRegionNames, false);
083  }
084
085  /**
086   * Like {@link Admin#unassign(byte[], boolean)} but 'raw' in that it can do more than one Region
087   * at a time -- good if many Regions to offline -- and it will schedule the assigns even in the
088   * case where Master is initializing (as long as the ProcedureExecutor is up). Does NOT call
089   * Coprocessor hooks.
090   * @param override           You need to add the override for case where a region has previously
091   *                           been bypassed. When a Procedure has been bypassed, a Procedure will
092   *                           have completed but no other Procedure will be able to make progress
093   *                           on the target entity (intentionally). This override flag will
094   *                           override this fencing mechanism.
095   * @param encodedRegionNames Region encoded names; e.g. 1588230740 is the hard-coded encoding for
096   *                           hbase:meta region and de00010733901a05f5a2a3a382e27dd4 is an example
097   *                           of what a random user-space encoded Region name looks like.
098   */
099  List<Long> unassigns(List<String> encodedRegionNames, boolean override) throws IOException;
100
101  default List<Long> unassigns(List<String> encodedRegionNames) throws IOException {
102    return unassigns(encodedRegionNames, false);
103  }
104
105  /**
106   * Bypass specified procedure and move it to completion. Procedure is marked completed but no
107   * actual work is done from the current state/step onwards. Parents of the procedure are also
108   * marked for bypass.
109   * @param pids      of procedures to complete.
110   * @param waitTime  wait time in ms for acquiring lock for a procedure
111   * @param override  if override set to true, we will bypass the procedure even if it is executing.
112   *                  This is for procedures which can't break out during execution (bugs?).
113   * @param recursive If set, if a parent procedure, we will find and bypass children and then the
114   *                  parent procedure (Dangerous but useful in case where child procedure has been
115   *                  'lost'). Does not always work. Experimental.
116   * @return true if procedure is marked for bypass successfully, false otherwise
117   */
118  List<Boolean> bypassProcedure(List<Long> pids, long waitTime, boolean override, boolean recursive)
119    throws IOException;
120
121  /**
122   * Use {@link #scheduleServerCrashProcedures(List)} instead.
123   * @deprecated since 2.2.1. Will removed in 3.0.0.
124   */
125  @Deprecated
126  default List<Long> scheduleServerCrashProcedure(List<HBaseProtos.ServerName> serverNames)
127    throws IOException {
128    return scheduleServerCrashProcedures(
129      serverNames.stream().map(ProtobufUtil::toServerName).collect(Collectors.toList()));
130  }
131
132  List<Long> scheduleServerCrashProcedures(List<ServerName> serverNames) throws IOException;
133
134  List<Long> scheduleSCPsForUnknownServers() throws IOException;
135
136  /**
137   * Request HBCK chore to run at master side.
138   * @return <code>true</code> if HBCK chore ran, <code>false</code> if HBCK chore already running
139   * @throws IOException if a remote or network exception occurs
140   */
141  boolean runHbckChore() throws IOException;
142
143  /**
144   * Fix Meta.
145   */
146  void fixMeta() throws IOException;
147}