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.mapreduce;
019
020import java.io.IOException;
021import java.text.ParseException;
022import java.text.SimpleDateFormat;
023import java.util.ArrayList;
024import java.util.Collections;
025import java.util.HashSet;
026import java.util.List;
027import java.util.Map;
028import java.util.Set;
029import java.util.TreeMap;
030import org.apache.hadoop.conf.Configuration;
031import org.apache.hadoop.conf.Configured;
032import org.apache.hadoop.fs.Path;
033import org.apache.hadoop.hbase.Cell;
034import org.apache.hadoop.hbase.CellUtil;
035import org.apache.hadoop.hbase.HBaseConfiguration;
036import org.apache.hadoop.hbase.PrivateCellUtil;
037import org.apache.hadoop.hbase.TableName;
038import org.apache.hadoop.hbase.client.Connection;
039import org.apache.hadoop.hbase.client.ConnectionFactory;
040import org.apache.hadoop.hbase.client.Delete;
041import org.apache.hadoop.hbase.client.Mutation;
042import org.apache.hadoop.hbase.client.Put;
043import org.apache.hadoop.hbase.client.RegionLocator;
044import org.apache.hadoop.hbase.client.Table;
045import org.apache.hadoop.hbase.io.ImmutableBytesWritable;
046import org.apache.hadoop.hbase.mapreduce.HFileOutputFormat2.TableInfo;
047import org.apache.hadoop.hbase.regionserver.wal.WALCellCodec;
048import org.apache.hadoop.hbase.util.Bytes;
049import org.apache.hadoop.hbase.util.EnvironmentEdgeManager;
050import org.apache.hadoop.hbase.util.MapReduceExtendedCell;
051import org.apache.hadoop.hbase.wal.WALEdit;
052import org.apache.hadoop.hbase.wal.WALKey;
053import org.apache.hadoop.mapreduce.Job;
054import org.apache.hadoop.mapreduce.Mapper;
055import org.apache.hadoop.mapreduce.lib.input.FileInputFormat;
056import org.apache.hadoop.mapreduce.lib.output.FileOutputFormat;
057import org.apache.hadoop.util.Tool;
058import org.apache.hadoop.util.ToolRunner;
059import org.apache.yetus.audience.InterfaceAudience;
060import org.slf4j.Logger;
061import org.slf4j.LoggerFactory;
062
063/**
064 * A tool to replay WAL files as a M/R job. The WAL can be replayed for a set of tables or all
065 * tables, and a time range can be provided (in milliseconds). The WAL is filtered to the passed set
066 * of tables and the output can optionally be mapped to another set of tables. WAL replay can also
067 * generate HFiles for later bulk importing, in that case the WAL is replayed for a single table
068 * only.
069 */
070@InterfaceAudience.Public
071public class WALPlayer extends Configured implements Tool {
072  private static final Logger LOG = LoggerFactory.getLogger(WALPlayer.class);
073  final static String NAME = "WALPlayer";
074  public final static String BULK_OUTPUT_CONF_KEY = "wal.bulk.output";
075  public final static String TABLES_KEY = "wal.input.tables";
076  public final static String TABLE_MAP_KEY = "wal.input.tablesmap";
077  public final static String INPUT_FILES_SEPARATOR_KEY = "wal.input.separator";
078  public final static String IGNORE_MISSING_FILES = "wal.input.ignore.missing.files";
079  public final static String MULTI_TABLES_SUPPORT = "wal.multi.tables.support";
080
081  protected static final String tableSeparator = ";";
082
083  private final static String JOB_NAME_CONF_KEY = "mapreduce.job.name";
084
085  public WALPlayer() {
086  }
087
088  protected WALPlayer(final Configuration c) {
089    super(c);
090  }
091
092  /**
093   * A mapper that just writes out KeyValues. This one can be used together with
094   * {@link CellSortReducer}
095   */
096  static class WALKeyValueMapper extends Mapper<WALKey, WALEdit, ImmutableBytesWritable, Cell> {
097    private Set<String> tableSet = new HashSet<String>();
098    private boolean multiTableSupport = false;
099
100    @Override
101    public void map(WALKey key, WALEdit value, Context context) throws IOException {
102      try {
103        // skip all other tables
104        TableName table = key.getTableName();
105        if (tableSet.contains(table.getNameAsString())) {
106          for (Cell cell : value.getCells()) {
107            if (WALEdit.isMetaEditFamily(cell)) {
108              continue;
109            }
110
111            // Set sequenceId from WALKey, since it is not included by WALCellCodec. The sequenceId
112            // on WALKey is the same value that was on the cells in the WALEdit. This enables
113            // CellSortReducer to use sequenceId to disambiguate duplicate cell timestamps.
114            // See HBASE-27649
115            PrivateCellUtil.setSequenceId(cell, key.getSequenceId());
116
117            byte[] outKey = multiTableSupport
118              ? Bytes.add(table.getName(), Bytes.toBytes(tableSeparator), CellUtil.cloneRow(cell))
119              : CellUtil.cloneRow(cell);
120            context.write(new ImmutableBytesWritable(outKey), new MapReduceExtendedCell(cell));
121          }
122        }
123      } catch (InterruptedException e) {
124        LOG.error("Interrupted while emitting Cell", e);
125        Thread.currentThread().interrupt();
126      }
127    }
128
129    @Override
130    public void setup(Context context) throws IOException {
131      Configuration conf = context.getConfiguration();
132      String[] tables = conf.getStrings(TABLES_KEY);
133      this.multiTableSupport = conf.getBoolean(MULTI_TABLES_SUPPORT, false);
134      Collections.addAll(tableSet, tables);
135    }
136  }
137
138  /**
139   * Enum for map metrics. Keep it out here rather than inside in the Map inner-class so we can find
140   * associated properties.
141   */
142  protected static enum Counter {
143    /** Number of aggregated writes */
144    PUTS,
145    /** Number of aggregated deletes */
146    DELETES,
147    CELLS_READ,
148    CELLS_WRITTEN,
149    WALEDITS
150  }
151
152  /**
153   * A mapper that writes out {@link Mutation} to be directly applied to a running HBase instance.
154   */
155  protected static class WALMapper
156    extends Mapper<WALKey, WALEdit, ImmutableBytesWritable, Mutation> {
157    private Map<TableName, TableName> tables = new TreeMap<>();
158
159    @Override
160    public void map(WALKey key, WALEdit value, Context context) throws IOException {
161      context.getCounter(Counter.WALEDITS).increment(1);
162      try {
163        if (tables.isEmpty() || tables.containsKey(key.getTableName())) {
164          TableName targetTable =
165            tables.isEmpty() ? key.getTableName() : tables.get(key.getTableName());
166          ImmutableBytesWritable tableOut = new ImmutableBytesWritable(targetTable.getName());
167          Put put = null;
168          Delete del = null;
169          Cell lastCell = null;
170          for (Cell cell : value.getCells()) {
171            context.getCounter(Counter.CELLS_READ).increment(1);
172            // Filtering WAL meta marker entries.
173            if (WALEdit.isMetaEditFamily(cell)) {
174              continue;
175            }
176            // Allow a subclass filter out this cell.
177            if (filter(context, cell)) {
178              // A WALEdit may contain multiple operations (HBASE-3584) and/or
179              // multiple rows (HBASE-5229).
180              // Aggregate as much as possible into a single Put/Delete
181              // operation before writing to the context.
182              if (
183                lastCell == null || lastCell.getTypeByte() != cell.getTypeByte()
184                  || !CellUtil.matchingRows(lastCell, cell)
185              ) {
186                // row or type changed, write out aggregate KVs.
187                if (put != null) {
188                  context.write(tableOut, put);
189                  context.getCounter(Counter.PUTS).increment(1);
190                }
191                if (del != null) {
192                  context.write(tableOut, del);
193                  context.getCounter(Counter.DELETES).increment(1);
194                }
195                if (CellUtil.isDelete(cell)) {
196                  del = new Delete(CellUtil.cloneRow(cell));
197                } else {
198                  put = new Put(CellUtil.cloneRow(cell));
199                }
200              }
201              if (CellUtil.isDelete(cell)) {
202                del.add(cell);
203              } else {
204                put.add(cell);
205              }
206              context.getCounter(Counter.CELLS_WRITTEN).increment(1);
207            }
208            lastCell = cell;
209          }
210          // write residual KVs
211          if (put != null) {
212            context.write(tableOut, put);
213            context.getCounter(Counter.PUTS).increment(1);
214          }
215          if (del != null) {
216            context.getCounter(Counter.DELETES).increment(1);
217            context.write(tableOut, del);
218          }
219        }
220      } catch (InterruptedException e) {
221        LOG.error("Interrupted while writing results", e);
222        Thread.currentThread().interrupt();
223      }
224    }
225
226    protected boolean filter(Context context, final Cell cell) {
227      return true;
228    }
229
230    @Override
231    protected void
232      cleanup(Mapper<WALKey, WALEdit, ImmutableBytesWritable, Mutation>.Context context)
233        throws IOException, InterruptedException {
234      super.cleanup(context);
235    }
236
237    @SuppressWarnings("checkstyle:EmptyBlock")
238    @Override
239    public void setup(Context context) throws IOException {
240      String[] tableMap = context.getConfiguration().getStrings(TABLE_MAP_KEY);
241      String[] tablesToUse = context.getConfiguration().getStrings(TABLES_KEY);
242      if (tableMap == null) {
243        tableMap = tablesToUse;
244      }
245      if (tablesToUse == null) {
246        // Then user wants all tables.
247      } else if (tablesToUse.length != tableMap.length) {
248        // this can only happen when WALMapper is used directly by a class other than WALPlayer
249        throw new IOException("Incorrect table mapping specified .");
250      }
251      int i = 0;
252      if (tablesToUse != null) {
253        for (String table : tablesToUse) {
254          tables.put(TableName.valueOf(table), TableName.valueOf(tableMap[i++]));
255        }
256      }
257    }
258  }
259
260  void setupTime(Configuration conf, String option) throws IOException {
261    String val = conf.get(option);
262    if (null == val) {
263      return;
264    }
265    long ms;
266    try {
267      // first try to parse in user friendly form
268      ms = new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss.SS").parse(val).getTime();
269    } catch (ParseException pe) {
270      try {
271        // then see if just a number of ms's was specified
272        ms = Long.parseLong(val);
273      } catch (NumberFormatException nfe) {
274        throw new IOException(
275          option + " must be specified either in the form 2001-02-20T16:35:06.99 "
276            + "or as number of milliseconds");
277      }
278    }
279    conf.setLong(option, ms);
280  }
281
282  /**
283   * Sets up the actual job.
284   * @param args The command line parameters.
285   * @return The newly created job.
286   * @throws IOException When setting up the job fails.
287   */
288  public Job createSubmittableJob(String[] args) throws IOException {
289    Configuration conf = getConf();
290    setupTime(conf, WALInputFormat.START_TIME_KEY);
291    setupTime(conf, WALInputFormat.END_TIME_KEY);
292    String inputDirs = args[0];
293    String[] tables = args.length == 1 ? new String[] {} : args[1].split(",");
294    String[] tableMap;
295    if (args.length > 2) {
296      tableMap = args[2].split(",");
297      if (tableMap.length != tables.length) {
298        throw new IOException("The same number of tables and mapping must be provided.");
299      }
300    } else {
301      // if no mapping is specified, map each table to itself
302      tableMap = tables;
303    }
304    conf.setStrings(TABLES_KEY, tables);
305    conf.setStrings(TABLE_MAP_KEY, tableMap);
306    conf.set(FileInputFormat.INPUT_DIR, inputDirs);
307    Job job = Job.getInstance(conf,
308      conf.get(JOB_NAME_CONF_KEY, NAME + "_" + EnvironmentEdgeManager.currentTime()));
309    job.setJarByClass(WALPlayer.class);
310
311    job.setInputFormatClass(WALInputFormat.class);
312    job.setMapOutputKeyClass(ImmutableBytesWritable.class);
313
314    String hfileOutPath = conf.get(BULK_OUTPUT_CONF_KEY);
315    if (hfileOutPath != null) {
316      LOG.debug("add incremental job :" + hfileOutPath + " from " + inputDirs);
317
318      // WALPlayer needs ExtendedCellSerialization so that sequenceId can be propagated when
319      // sorting cells in CellSortReducer
320      job.getConfiguration().setBoolean(HFileOutputFormat2.EXTENDED_CELL_SERIALIZATION_ENABLED_KEY,
321        true);
322
323      // the bulk HFile case
324      List<TableName> tableNames = getTableNameList(tables);
325
326      job.setMapperClass(WALKeyValueMapper.class);
327      job.setReducerClass(CellSortReducer.class);
328      Path outputDir = new Path(hfileOutPath);
329      FileOutputFormat.setOutputPath(job, outputDir);
330      job.setMapOutputValueClass(MapReduceExtendedCell.class);
331      try (Connection conn = ConnectionFactory.createConnection(conf);) {
332        List<TableInfo> tableInfoList = new ArrayList<TableInfo>();
333        for (TableName tableName : tableNames) {
334          Table table = conn.getTable(tableName);
335          RegionLocator regionLocator = conn.getRegionLocator(tableName);
336          tableInfoList.add(new TableInfo(table.getDescriptor(), regionLocator));
337        }
338        MultiTableHFileOutputFormat.configureIncrementalLoad(job, tableInfoList);
339      }
340      TableMapReduceUtil.addDependencyJarsForClasses(job.getConfiguration(),
341        org.apache.hbase.thirdparty.com.google.common.base.Preconditions.class);
342    } else {
343      // output to live cluster
344      job.setMapperClass(WALMapper.class);
345      job.setOutputFormatClass(MultiTableOutputFormat.class);
346      TableMapReduceUtil.addDependencyJars(job);
347      TableMapReduceUtil.initCredentials(job);
348      // No reducers.
349      job.setNumReduceTasks(0);
350    }
351    String codecCls = WALCellCodec.getWALCellCodecClass(conf).getName();
352    try {
353      TableMapReduceUtil.addDependencyJarsForClasses(job.getConfiguration(),
354        Class.forName(codecCls));
355    } catch (Exception e) {
356      throw new IOException("Cannot determine wal codec class " + codecCls, e);
357    }
358    return job;
359  }
360
361  private List<TableName> getTableNameList(String[] tables) {
362    List<TableName> list = new ArrayList<TableName>();
363    for (String name : tables) {
364      list.add(TableName.valueOf(name));
365    }
366    return list;
367  }
368
369  /**
370   * Print usage
371   * @param errorMsg Error message. Can be null.
372   */
373  private void usage(final String errorMsg) {
374    if (errorMsg != null && errorMsg.length() > 0) {
375      System.err.println("ERROR: " + errorMsg);
376    }
377    System.err.println("Usage: " + NAME + " [options] <WAL inputdir> [<tables> <tableMappings>]");
378    System.err.println(" <WAL inputdir>   directory of WALs to replay.");
379    System.err.println(" <tables>         comma separated list of tables. If no tables specified,");
380    System.err.println("                  all are imported (even hbase:meta if present).");
381    System.err.println(
382      " <tableMappings>  WAL entries can be mapped to a new set of tables by " + "passing");
383    System.err
384      .println("                  <tableMappings>, a comma separated list of target " + "tables.");
385    System.err
386      .println("                  If specified, each table in <tables> must have a " + "mapping.");
387    System.err.println("To generate HFiles to bulk load instead of loading HBase directly, pass:");
388    System.err.println(" -D" + BULK_OUTPUT_CONF_KEY + "=/path/for/output");
389    System.err.println(" Only one table can be specified, and no mapping allowed!");
390    System.err.println("To specify a time range, pass:");
391    System.err.println(" -D" + WALInputFormat.START_TIME_KEY + "=[date|ms]");
392    System.err.println(" -D" + WALInputFormat.END_TIME_KEY + "=[date|ms]");
393    System.err.println(" The start and the end date of timerange (inclusive). The dates can be");
394    System.err
395      .println(" expressed in milliseconds-since-epoch or yyyy-MM-dd'T'HH:mm:ss.SS " + "format.");
396    System.err.println(" E.g. 1234567890120 or 2009-02-13T23:32:30.12");
397    System.err.println("Other options:");
398    System.err.println(" -D" + JOB_NAME_CONF_KEY + "=jobName");
399    System.err.println(" Use the specified mapreduce job name for the wal player");
400    System.err.println(" -Dwal.input.separator=' '");
401    System.err.println(" Change WAL filename separator (WAL dir names use default ','.)");
402    System.err.println("For performance also consider the following options:\n"
403      + "  -Dmapreduce.map.speculative=false\n" + "  -Dmapreduce.reduce.speculative=false");
404  }
405
406  /**
407   * Main entry point.
408   * @param args The command line parameters.
409   * @throws Exception When running the job fails.
410   */
411  public static void main(String[] args) throws Exception {
412    int ret = ToolRunner.run(new WALPlayer(HBaseConfiguration.create()), args);
413    System.exit(ret);
414  }
415
416  @Override
417  public int run(String[] args) throws Exception {
418    if (args.length < 1) {
419      usage("Wrong number of arguments: " + args.length);
420      System.exit(-1);
421    }
422    Job job = createSubmittableJob(args);
423    return job.waitForCompletion(true) ? 0 : 1;
424  }
425}