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 Connection#getHbck()} and call
036 * {@link #close()} when done.
037 * <p>WARNING: the below methods can damage the cluster. It may leave the cluster in an
038 * indeterminate state, e.g. region not assigned, or some hdfs files left behind. After running
039 * any of the below, operators may have to do some clean up on hdfs or schedule some assign
040 * procedures to get regions back online. DO AT YOUR OWN RISK. For experienced users only.
041 *
042 * @see ConnectionFactory
043 * @since 2.0.2, 2.1.1
044 */
045@InterfaceAudience.LimitedPrivate(HBaseInterfaceAudience.HBCK)
046public interface Hbck extends Abortable, Closeable {
047  /**
048   * Update table state in Meta only. No procedures are submitted to open/assign or
049   * close/unassign regions of the table.
050   * @param state table state
051   * @return previous state of the table in Meta
052   */
053  TableState setTableStateInMeta(TableState state) throws IOException;
054
055  /**
056   * Update region state in Meta only. No procedures are submitted to manipulate the given region or
057   * any other region from same table.
058   * @param nameOrEncodedName2State list of all region states to be updated in meta
059   * @return previous state of the region in Meta
060   */
061  Map<String, RegionState.State>
062    setRegionStateInMeta(Map<String, RegionState.State> nameOrEncodedName2State) throws IOException;
063
064  /**
065   * Like {@link Admin#assign(byte[])} but 'raw' in that it can do more than one Region at a time
066   * -- good if many Regions to online -- and it will schedule the assigns even in the case where
067   * Master is initializing (as long as the ProcedureExecutor is up). Does NOT call Coprocessor
068   * hooks.
069   * @param override You need to add the override for case where a region has previously been
070   *              bypassed. When a Procedure has been bypassed, a Procedure will have completed
071   *              but no other Procedure will be able to make progress on the target entity
072   *              (intentionally). This override flag will override this fencing mechanism.
073   * @param encodedRegionNames Region encoded names; e.g. 1588230740 is the hard-coded encoding
074   *                           for hbase:meta region and de00010733901a05f5a2a3a382e27dd4 is an
075   *                           example of what a random user-space encoded Region name looks like.
076   */
077  List<Long> assigns(List<String> encodedRegionNames, boolean override) throws IOException;
078
079  default List<Long> assigns(List<String> encodedRegionNames) throws IOException {
080    return assigns(encodedRegionNames, false);
081  }
082
083  /**
084   * Like {@link Admin#unassign(byte[], boolean)} but 'raw' in that it can do more than one Region
085   * at a time -- good if many Regions to offline -- and it will schedule the assigns even in the
086   * case where Master is initializing (as long as the ProcedureExecutor is up). Does NOT call
087   * Coprocessor hooks.
088   * @param override You need to add the override for case where a region has previously been
089   *              bypassed. When a Procedure has been bypassed, a Procedure will have completed
090   *              but no other Procedure will be able to make progress on the target entity
091   *              (intentionally). This override flag will override this fencing mechanism.
092   * @param encodedRegionNames Region encoded names; e.g. 1588230740 is the hard-coded encoding
093   *                           for hbase:meta region and de00010733901a05f5a2a3a382e27dd4 is an
094   *                           example of what a random user-space encoded Region name looks like.
095   */
096  List<Long> unassigns(List<String> encodedRegionNames, boolean override) throws IOException;
097
098  default List<Long> unassigns(List<String> encodedRegionNames) throws IOException {
099    return unassigns(encodedRegionNames, false);
100  }
101
102  /**
103   * Bypass specified procedure and move it to completion. Procedure is marked completed but
104   * no actual work is done from the current state/step onwards. Parents of the procedure are
105   * also marked for bypass.
106   *
107   * @param pids of procedures to complete.
108   * @param waitTime wait time in ms for acquiring lock for a procedure
109   * @param override if override set to true, we will bypass the procedure even if it is executing.
110   *   This is for procedures which can't break out during execution (bugs?).
111   * @param recursive If set, if a parent procedure, we will find and bypass children and then
112   *   the parent procedure (Dangerous but useful in case where child procedure has been 'lost').
113   *   Does not always work. Experimental.
114   * @return true if procedure is marked for bypass successfully, false otherwise
115   */
116  List<Boolean> bypassProcedure(List<Long> pids, long waitTime, boolean override, boolean recursive)
117      throws IOException;
118
119  /**
120   * Use {@link #scheduleServerCrashProcedures(List)} instead.
121   * @deprecated since 2.2.1. Will removed in 3.0.0.
122   */
123  @Deprecated
124  default List<Long> scheduleServerCrashProcedure(List<HBaseProtos.ServerName> serverNames)
125      throws IOException {
126    return scheduleServerCrashProcedures(
127        serverNames.stream().map(ProtobufUtil::toServerName).collect(Collectors.toList()));
128  }
129
130  List<Long> scheduleServerCrashProcedures(List<ServerName> serverNames) throws IOException;
131
132  /**
133   * Request HBCK chore to run at master side.
134   *
135   * @return <code>true</code> if HBCK chore ran, <code>false</code> if HBCK chore already running
136   * @throws IOException if a remote or network exception occurs
137   */
138  boolean runHbckChore() throws IOException;
139
140  /**
141   * Fix Meta.
142   */
143  void fixMeta() throws IOException;
144}