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.regionserver;
019
020import java.io.IOException;
021import org.apache.hadoop.conf.Configuration;
022import org.apache.hadoop.fs.FileSystem;
023import org.apache.hadoop.fs.Path;
024import org.apache.hadoop.hbase.HBaseClassTestRule;
025import org.apache.hadoop.hbase.HBaseTestingUtil;
026import org.apache.hadoop.hbase.HConstants;
027import org.apache.hadoop.hbase.KeyValue;
028import org.apache.hadoop.hbase.ServerName;
029import org.apache.hadoop.hbase.TableName;
030import org.apache.hadoop.hbase.client.ColumnFamilyDescriptorBuilder;
031import org.apache.hadoop.hbase.client.RegionInfo;
032import org.apache.hadoop.hbase.client.RegionInfoBuilder;
033import org.apache.hadoop.hbase.client.TableDescriptor;
034import org.apache.hadoop.hbase.client.TableDescriptorBuilder;
035import org.apache.hadoop.hbase.monitoring.TaskMonitor;
036import org.apache.hadoop.hbase.testclassification.RegionServerTests;
037import org.apache.hadoop.hbase.testclassification.SmallTests;
038import org.apache.hadoop.hbase.util.Bytes;
039import org.apache.hadoop.hbase.util.CancelableProgressable;
040import org.apache.hadoop.hbase.util.CommonFSUtils;
041import org.apache.hadoop.hbase.wal.WAL;
042import org.apache.hadoop.hbase.wal.WALEdit;
043import org.apache.hadoop.hbase.wal.WALFactory;
044import org.apache.hadoop.hbase.wal.WALKeyImpl;
045import org.apache.hadoop.hbase.wal.WALProvider;
046import org.apache.hadoop.hbase.wal.WALSplitUtil;
047import org.junit.After;
048import org.junit.Assert;
049import org.junit.Before;
050import org.junit.ClassRule;
051import org.junit.Rule;
052import org.junit.Test;
053import org.junit.experimental.categories.Category;
054import org.junit.rules.TestName;
055import org.mockito.Mockito;
056import org.slf4j.Logger;
057import org.slf4j.LoggerFactory;
058
059/**
060 * HBASE-21031 If replay edits fails, we need to make sure memstore is rollbacked And if MSLAB is
061 * used, all chunk is released too.
062 */
063@Category({ RegionServerTests.class, SmallTests.class })
064public class TestRecoveredEditsReplayAndAbort {
065  @ClassRule
066  public static final HBaseClassTestRule CLASS_RULE =
067    HBaseClassTestRule.forClass(TestRecoveredEditsReplayAndAbort.class);
068
069  private static final Logger LOG = LoggerFactory.getLogger(TestRecoveredEditsReplayAndAbort.class);
070
071  protected final byte[] row = Bytes.toBytes("rowA");
072
073  protected final static byte[] fam1 = Bytes.toBytes("colfamily11");
074
075  @Rule
076  public TestName name = new TestName();
077
078  // Test names
079  protected TableName tableName;
080  protected String method;
081
082  protected static HBaseTestingUtil TEST_UTIL;
083  public static Configuration CONF;
084  private HRegion region = null;
085
086  @Before
087  public void setup() throws IOException {
088    TEST_UTIL = new HBaseTestingUtil();
089    CONF = TEST_UTIL.getConfiguration();
090    method = name.getMethodName();
091    tableName = TableName.valueOf(method);
092  }
093
094  @After
095  public void tearDown() throws Exception {
096    LOG.info("Cleaning test directory: " + TEST_UTIL.getDataTestDir());
097    TEST_UTIL.cleanupTestDir();
098  }
099
100  @Test
101  public void test() throws Exception {
102    // set flush size to 10MB
103    CONF.setInt("hbase.hregion.memstore.flush.size", 1024 * 1024 * 10);
104    // set the report interval to a very small value
105    CONF.setInt("hbase.hstore.report.interval.edits", 1);
106    CONF.setInt("hbase.hstore.report.period", 0);
107    // mock a RegionServerServices
108    final RegionServerAccounting rsAccounting = new RegionServerAccounting(CONF);
109    RegionServerServices rs = Mockito.mock(RegionServerServices.class);
110    ChunkCreator.initialize(MemStoreLAB.CHUNK_SIZE_DEFAULT, false, 0, 0, 0, null,
111      MemStoreLAB.INDEX_CHUNK_SIZE_PERCENTAGE_DEFAULT);
112    Mockito.when(rs.getRegionServerAccounting()).thenReturn(rsAccounting);
113    Mockito.when(rs.isAborted()).thenReturn(false);
114    Mockito.when(rs.getNonceManager()).thenReturn(null);
115    Mockito.when(rs.getServerName()).thenReturn(ServerName.valueOf("test", 0, 111));
116    Mockito.when(rs.getConfiguration()).thenReturn(CONF);
117    // create a region
118    TableName testTable = TableName.valueOf("testRecoveredEidtsReplayAndAbort");
119    TableDescriptor htd = TableDescriptorBuilder.newBuilder(testTable)
120      .setColumnFamily(ColumnFamilyDescriptorBuilder.newBuilder(fam1).build()).build();
121    RegionInfo info = RegionInfoBuilder.newBuilder(htd.getTableName()).build();
122    Path logDir = TEST_UTIL.getDataTestDirOnTestFS("TestRecoveredEidtsReplayAndAbort.log");
123    final WAL wal = HBaseTestingUtil.createWal(CONF, logDir, info);
124    Path rootDir = TEST_UTIL.getDataTestDir();
125    Path tableDir = CommonFSUtils.getTableDir(rootDir, info.getTable());
126    HRegionFileSystem.createRegionOnFileSystem(CONF, TEST_UTIL.getTestFileSystem(), tableDir, info);
127    region = HRegion.newHRegion(tableDir, wal, TEST_UTIL.getTestFileSystem(), CONF, info, htd, rs);
128    // create some recovered.edits
129    final WALFactory wals = new WALFactory(CONF, method);
130    try {
131      Path regiondir = region.getRegionFileSystem().getRegionDir();
132      FileSystem fs = region.getRegionFileSystem().getFileSystem();
133      byte[] regionName = region.getRegionInfo().getEncodedNameAsBytes();
134
135      Path recoveredEditsDir = WALSplitUtil.getRegionDirRecoveredEditsDir(regiondir);
136      long maxSeqId = 1200;
137      long minSeqId = 1000;
138      long totalEdits = maxSeqId - minSeqId;
139      for (long i = minSeqId; i <= maxSeqId; i += 100) {
140        Path recoveredEdits = new Path(recoveredEditsDir, String.format("%019d", i));
141        LOG.info("Begin to write recovered.edits : " + recoveredEdits);
142        fs.create(recoveredEdits);
143        WALProvider.Writer writer = wals.createRecoveredEditsWriter(fs, recoveredEdits);
144        for (long j = i; j < i + 100; j++) {
145          long time = System.nanoTime();
146          WALEdit edit = new WALEdit();
147          // 200KB kv
148          byte[] value = new byte[200 * 1024];
149          Bytes.random(value);
150          edit.add(new KeyValue(row, fam1, Bytes.toBytes(j), time, KeyValue.Type.Put, value));
151          writer.append(new WAL.Entry(
152            new WALKeyImpl(regionName, tableName, j, time, HConstants.DEFAULT_CLUSTER_ID), edit));
153        }
154        writer.close();
155      }
156      TaskMonitor.get().createStatus(method);
157      // try to replay the edits
158      try {
159        region.initialize(new CancelableProgressable() {
160          private long replayedEdits = 0;
161
162          @Override
163          public boolean progress() {
164            replayedEdits++;
165            // during replay, rsAccounting should align with global memstore, because
166            // there is only one memstore here
167            Assert.assertEquals(rsAccounting.getGlobalMemStoreDataSize(),
168              region.getMemStoreDataSize());
169            Assert.assertEquals(rsAccounting.getGlobalMemStoreHeapSize(),
170              region.getMemStoreHeapSize());
171            Assert.assertEquals(rsAccounting.getGlobalMemStoreOffHeapSize(),
172              region.getMemStoreOffHeapSize());
173            // abort the replay before finishing, leaving some edits in the memory
174            return replayedEdits < totalEdits - 10;
175          }
176        });
177        Assert.fail("Should not reach here");
178      } catch (IOException t) {
179        LOG.info("Current memstore: " + region.getMemStoreDataSize() + ", "
180          + region.getMemStoreHeapSize() + ", " + region.getMemStoreOffHeapSize());
181      }
182      // After aborting replay, there should be no data in the memory
183      Assert.assertEquals(0, rsAccounting.getGlobalMemStoreDataSize());
184      Assert.assertEquals(0, region.getMemStoreDataSize());
185      // All the chunk in the MSLAB should be recycled, otherwise, there might be
186      // a memory leak.
187      Assert.assertEquals(0, ChunkCreator.getInstance().numberOfMappedChunks());
188    } finally {
189      HBaseTestingUtil.closeRegionAndWAL(this.region);
190      this.region = null;
191      wals.close();
192    }
193  }
194}