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