HBASE-27649 WALPlayer does not properly dedupe overridden cell versions (#5058)

Signed-off-by: Duo Zhang <zhangduo@apache.org>

hbase-mapreduce/src/main/java/org/apache/hadoop/hbase/mapreduce/ExtendedCellSerialization.java

@@ -0,0 +1,101 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.hadoop.hbase.mapreduce;
+
+import java.io.DataInputStream;
+import java.io.DataOutputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.io.OutputStream;
+import org.apache.hadoop.hbase.ExtendedCell;
+import org.apache.hadoop.hbase.KeyValue;
+import org.apache.hadoop.hbase.KeyValueUtil;
+import org.apache.hadoop.hbase.PrivateCellUtil;
+import org.apache.hadoop.hbase.util.Bytes;
+import org.apache.hadoop.io.serializer.Deserializer;
+import org.apache.hadoop.io.serializer.Serialization;
+import org.apache.hadoop.io.serializer.Serializer;
+import org.apache.yetus.audience.InterfaceAudience;
+
+/**
+ * Similar to CellSerialization, but includes the sequenceId from an ExtendedCell. This is necessary
+ * so that CellSortReducer can sort by sequenceId, if applicable. Note that these two serializations
+ * are not compatible -- data serialized by CellSerialization cannot be deserialized with
+ * ExtendedCellSerialization and vice versa. This is ok for {@link HFileOutputFormat2} because the
+ * serialization is not actually used for the actual written HFiles, just intermediate data (between
+ * mapper and reducer of a single job).
+ */
+@InterfaceAudience.Private
+public class ExtendedCellSerialization implements Serialization<ExtendedCell> {
+  @Override
+  public boolean accept(Class<?> c) {
+    return ExtendedCell.class.isAssignableFrom(c);
+  }
+
+  @Override
+  public ExtendedCellDeserializer getDeserializer(Class<ExtendedCell> t) {
+    return new ExtendedCellDeserializer();
+  }
+
+  @Override
+  public ExtendedCellSerializer getSerializer(Class<ExtendedCell> c) {
+    return new ExtendedCellSerializer();
+  }
+
+  public static class ExtendedCellDeserializer implements Deserializer<ExtendedCell> {
+    private DataInputStream dis;
+
+    @Override
+    public void close() throws IOException {
+      this.dis.close();
+    }
+
+    @Override
+    public KeyValue deserialize(ExtendedCell ignore) throws IOException {
+      KeyValue kv = KeyValueUtil.create(this.dis);
+      PrivateCellUtil.setSequenceId(kv, this.dis.readLong());
+      return kv;
+    }
+
+    @Override
+    public void open(InputStream is) throws IOException {
+      this.dis = new DataInputStream(is);
+    }
+  }
+
+  public static class ExtendedCellSerializer implements Serializer<ExtendedCell> {
+    private DataOutputStream dos;
+
+    @Override
+    public void close() throws IOException {
+      this.dos.close();
+    }
+
+    @Override
+    public void open(OutputStream os) throws IOException {
+      this.dos = new DataOutputStream(os);
+    }
+
+    @Override
+    public void serialize(ExtendedCell kv) throws IOException {
+      dos.writeInt(PrivateCellUtil.estimatedSerializedSizeOf(kv) - Bytes.SIZEOF_INT);
+      PrivateCellUtil.writeCell(kv, dos, true);
+      dos.writeLong(kv.getSequenceId());
+    }
+  }
+}

hbase-mapreduce/src/main/java/org/apache/hadoop/hbase/mapreduce/HFileOutputFormat2.java

@@ -30,6 +30,7 @@ import java.net.URLEncoder;
 import java.nio.charset.Charset;
 import java.util.ArrayList;
 import java.util.Arrays;
+import java.util.Collections;
 import java.util.List;
 import java.util.Map;
 import java.util.Map.Entry;
@@ -159,6 +160,15 @@ public class HFileOutputFormat2 extends FileOutputFormat<ImmutableBytesWritable,
   static final String MULTI_TABLE_HFILEOUTPUTFORMAT_CONF_KEY =
     "hbase.mapreduce.use.multi.table.hfileoutputformat";
 
+  /**
+   * ExtendedCell and ExtendedCellSerialization are InterfaceAudience.Private. We expose this config
+   * package-private for internal usage for jobs like WALPlayer which need to use features of
+   * ExtendedCell.
+   */
+  static final String EXTENDED_CELL_SERIALIZATION_ENABLED_KEY =
+    "hbase.mapreduce.hfileoutputformat.extendedcell.enabled";
+  static final boolean EXTENDED_CELL_SERIALIZATION_ENABLED_DEFULT = false;
+
   public static final String REMOTE_CLUSTER_CONF_PREFIX = "hbase.hfileoutputformat.remote.cluster.";
   public static final String REMOTE_CLUSTER_ZOOKEEPER_QUORUM_CONF_KEY =
     REMOTE_CLUSTER_CONF_PREFIX + "zookeeper.quorum";
@@ -619,9 +629,7 @@ public class HFileOutputFormat2 extends FileOutputFormat<ImmutableBytesWritable,
       LOG.warn("Unknown map output value type:" + job.getMapOutputValueClass());
     }
 
-    conf.setStrings("io.serializations", conf.get("io.serializations"),
-      MutationSerialization.class.getName(), ResultSerialization.class.getName(),
-      CellSerialization.class.getName());
+    mergeSerializations(conf);
 
     if (conf.getBoolean(LOCALITY_SENSITIVE_CONF_KEY, DEFAULT_LOCALITY_SENSITIVE)) {
       LOG.info("bulkload locality sensitive enabled");
@@ -670,6 +678,33 @@ public class HFileOutputFormat2 extends FileOutputFormat<ImmutableBytesWritable,
     LOG.info("Incremental output configured for tables: " + StringUtils.join(allTableNames, ","));
   }
 
+  private static void mergeSerializations(Configuration conf) {
+    List<String> serializations = new ArrayList<>();
+
+    // add any existing values that have been set
+    String[] existing = conf.getStrings("io.serializations");
+    if (existing != null) {
+      Collections.addAll(serializations, existing);
+    }
+
+    serializations.add(MutationSerialization.class.getName());
+    serializations.add(ResultSerialization.class.getName());
+
+    // Add ExtendedCellSerialization, if configured. Order matters here. Hadoop's
+    // SerializationFactory runs through serializations in the order they are registered.
+    // We want to register ExtendedCellSerialization before CellSerialization because both
+    // work for ExtendedCells but only ExtendedCellSerialization handles them properly.
+    if (
+      conf.getBoolean(EXTENDED_CELL_SERIALIZATION_ENABLED_KEY,
+        EXTENDED_CELL_SERIALIZATION_ENABLED_DEFULT)
+    ) {
+      serializations.add(ExtendedCellSerialization.class.getName());
+    }
+    serializations.add(CellSerialization.class.getName());
+
+    conf.setStrings("io.serializations", serializations.toArray(new String[0]));
+  }
+
   public static void configureIncrementalLoadMap(Job job, TableDescriptor tableDescriptor)
     throws IOException {
     Configuration conf = job.getConfiguration();

hbase-mapreduce/src/main/java/org/apache/hadoop/hbase/mapreduce/WALPlayer.java

@@ -32,6 +32,7 @@ import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.hbase.Cell;
 import org.apache.hadoop.hbase.CellUtil;
 import org.apache.hadoop.hbase.HBaseConfiguration;
+import org.apache.hadoop.hbase.PrivateCellUtil;
 import org.apache.hadoop.hbase.TableName;
 import org.apache.hadoop.hbase.client.Connection;
 import org.apache.hadoop.hbase.client.ConnectionFactory;
@@ -105,6 +106,13 @@ public class WALPlayer extends Configured implements Tool {
             if (WALEdit.isMetaEditFamily(cell)) {
               continue;
             }
+
+            // Set sequenceId from WALKey, since it is not included by WALCellCodec. The sequenceId
+            // on WALKey is the same value that was on the cells in the WALEdit. This enables
+            // CellSortReducer to use sequenceId to disambiguate duplicate cell timestamps.
+            // See HBASE-27649
+            PrivateCellUtil.setSequenceId(cell, key.getSequenceId());
+
             byte[] outKey = multiTableSupport
               ? Bytes.add(table.getName(), Bytes.toBytes(tableSeparator), CellUtil.cloneRow(cell))
               : CellUtil.cloneRow(cell);
@@ -308,6 +316,11 @@ public class WALPlayer extends Configured implements Tool {
     if (hfileOutPath != null) {
       LOG.debug("add incremental job :" + hfileOutPath + " from " + inputDirs);
 
+      // WALPlayer needs ExtendedCellSerialization so that sequenceId can be propagated when
+      // sorting cells in CellSortReducer
+      job.getConfiguration().setBoolean(HFileOutputFormat2.EXTENDED_CELL_SERIALIZATION_ENABLED_KEY,
+        true);
+
       // the bulk HFile case
       List<TableName> tableNames = getTableNameList(tables);
 

hbase-mapreduce/src/test/java/org/apache/hadoop/hbase/mapreduce/TestWALPlayer.java

@@ -17,6 +17,10 @@
  */
 package org.apache.hadoop.hbase.mapreduce;
 
+import static org.hamcrest.CoreMatchers.equalTo;
+import static org.hamcrest.CoreMatchers.notNullValue;
+import static org.hamcrest.CoreMatchers.nullValue;
+import static org.hamcrest.MatcherAssert.assertThat;
 import static org.junit.Assert.assertEquals;
 import static org.junit.Assert.assertTrue;
 import static org.junit.Assert.fail;
@@ -29,6 +33,7 @@ import java.io.ByteArrayOutputStream;
 import java.io.File;
 import java.io.PrintStream;
 import java.util.ArrayList;
+import java.util.concurrent.ThreadLocalRandom;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.Path;
@@ -50,8 +55,10 @@ import org.apache.hadoop.hbase.mapreduce.WALPlayer.WALKeyValueMapper;
 import org.apache.hadoop.hbase.regionserver.TestRecoveredEdits;
 import org.apache.hadoop.hbase.testclassification.LargeTests;
 import org.apache.hadoop.hbase.testclassification.MapReduceTests;
+import org.apache.hadoop.hbase.tool.BulkLoadHFiles;
 import org.apache.hadoop.hbase.util.Bytes;
 import org.apache.hadoop.hbase.util.CommonFSUtils;
+import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
 import org.apache.hadoop.hbase.util.LauncherSecurityManager;
 import org.apache.hadoop.hbase.util.MapReduceExtendedCell;
 import org.apache.hadoop.hbase.wal.WAL;
@@ -131,6 +138,80 @@ public class TestWALPlayer {
     assertTrue(TEST_UTIL.countRows(tn) > 0);
   }
 
+  /**
+   * Tests that when you write multiple cells with the same timestamp they are properly sorted by
+   * their sequenceId in WALPlayer/CellSortReducer so that the correct one wins when querying from
+   * the resulting bulkloaded HFiles. See HBASE-27649
+   */
+  @Test
+  public void testWALPlayerBulkLoadWithOverriddenTimestamps() throws Exception {
+    final TableName tableName = TableName.valueOf(name.getMethodName() + "1");
+    final byte[] family = Bytes.toBytes("family");
+    final byte[] column1 = Bytes.toBytes("c1");
+    final byte[] column2 = Bytes.toBytes("c2");
+    final byte[] row = Bytes.toBytes("row");
+    Table table = TEST_UTIL.createTable(tableName, family);
+
+    long now = EnvironmentEdgeManager.currentTime();
+    // put a row into the first table
+    Put p = new Put(row);
+    p.addColumn(family, column1, now, column1);
+    p.addColumn(family, column2, now, column2);
+
+    table.put(p);
+
+    byte[] lastVal = null;
+
+    for (int i = 0; i < 50; i++) {
+      lastVal = Bytes.toBytes(ThreadLocalRandom.current().nextLong());
+      p = new Put(row);
+      p.addColumn(family, column1, now, lastVal);
+
+      table.put(p);
+
+      // wal rolling is necessary to trigger the bug. otherwise no sorting
+      // needs to occur in the reducer because it's all sorted and coming from a single file.
+      if (i % 10 == 0) {
+        WAL log = cluster.getRegionServer(0).getWAL(null);
+        log.rollWriter();
+      }
+    }
+
+    WAL log = cluster.getRegionServer(0).getWAL(null);
+    log.rollWriter();
+    String walInputDir = new Path(cluster.getMaster().getMasterFileSystem().getWALRootDir(),
+      HConstants.HREGION_LOGDIR_NAME).toString();
+
+    Configuration configuration = new Configuration(TEST_UTIL.getConfiguration());
+    String outPath = "/tmp/" + name.getMethodName();
+    configuration.set(WALPlayer.BULK_OUTPUT_CONF_KEY, outPath);
+    configuration.setBoolean(WALPlayer.MULTI_TABLES_SUPPORT, true);
+
+    WALPlayer player = new WALPlayer(configuration);
+    assertEquals(0, ToolRunner.run(configuration, player,
+      new String[] { walInputDir, tableName.getNameAsString() }));
+
+    Get g = new Get(row);
+    Result result = table.get(g);
+    byte[] value = CellUtil.cloneValue(result.getColumnLatestCell(family, column1));
+    assertThat(Bytes.toStringBinary(value), equalTo(Bytes.toStringBinary(lastVal)));
+
+    table = TEST_UTIL.truncateTable(tableName);
+    g = new Get(row);
+    result = table.get(g);
+    assertThat(result.listCells(), nullValue());
+
+    BulkLoadHFiles.create(configuration).bulkLoad(tableName,
+      new Path(outPath, tableName.getNamespaceAsString() + "/" + tableName.getNameAsString()));
+
+    g = new Get(row);
+    result = table.get(g);
+    value = CellUtil.cloneValue(result.getColumnLatestCell(family, column1));
+
+    assertThat(result.listCells(), notNullValue());
+    assertThat(Bytes.toStringBinary(value), equalTo(Bytes.toStringBinary(lastVal)));
+  }
+
   /**
    * Simple end-to-end test
    */