Asynchronous Hadoop MapRunner

AsyncMapRunner.java

/**
 * 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.mapred;

import java.io.IOException;
import java.util.concurrent.ArrayBlockingQueue;
import java.util.concurrent.BlockingQueue;
import java.util.concurrent.Callable;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.Future;

import org.apache.commons.logging.Log;
import org.apache.commons.logging.LogFactory;
import org.apache.hadoop.util.ReflectionUtils;

/**
 * Async map runner that separates I/O and CPU workloads into 2 threads. It can
 * be used instead of the default implementation to gain a good throughput.
 */
public class AsyncMapRunner<K1, V1, K2, V2> implements
    MapRunnable<K1, V1, K2, V2> {

  private static final Log LOG = LogFactory.getLog(AsyncMapRunner.class
      .getName());

  private static final int DEFAULT_QUEUE_SIZE = 1024;
  private BlockingQueue<KvPair<K1, V1>> queue;
  private final KvPair<K1, V1> FINISH_TOKEN = new KvPair<K1, V1>();

  private Mapper<K1, V1, K2, V2> mapper;
  private ExecutorService executorService;
  private boolean incrProcCount;

  static class KvPair<K1, V1> {
    K1 key;
    V1 value;

    // used for finished token
    KvPair() {
    }

    KvPair(K1 key, V1 value) {
      this.key = key;
      this.value = value;
    }
  }

  @Override
  @SuppressWarnings("unchecked")
  public void configure(JobConf job) {
    int size =
        job.getInt("mapred.map.asyncmaprunner.queue.size", DEFAULT_QUEUE_SIZE);
    queue = new ArrayBlockingQueue<KvPair<K1, V1>>(size);
    if (LOG.isDebugEnabled()) {
      LOG.debug("Configuring jobConf " + job.getJobName() + " to use a " + size
          + " length queue");
    }

    this.mapper = ReflectionUtils.newInstance(job.getMapperClass(), job);
    // increment processed counter only if skipping feature is enabled
    this.incrProcCount =
        SkipBadRecords.getMapperMaxSkipRecords(job) > 0
            && SkipBadRecords.getAutoIncrMapperProcCount(job);
    executorService = Executors.newSingleThreadExecutor();
  }

  @Override
  public void run(RecordReader<K1, V1> input,
      final OutputCollector<K2, V2> output, final Reporter reporter)
      throws IOException {
    try {
      Future<Void> future = executorService.submit(new Callable<Void>() {
        public Void call() throws IOException {
          while (true) {
            try {
              KvPair<K1, V1> pair = queue.take();
              if (pair == FINISH_TOKEN) {
                // finished
                break;
              }
              mapper.map(pair.key, pair.value, output, reporter);
              if (incrProcCount) {
                reporter.incrCounter(SkipBadRecords.COUNTER_GROUP,
                    SkipBadRecords.COUNTER_MAP_PROCESSED_RECORDS, 1);
              }
            } catch (InterruptedException ex) {
              // task is cancelled or executor service is shutdown
              break;
            }
          }
          return null;
        }
      });

      // allocate key & value instances that are re-used for all entries
      K1 key = input.createKey();
      V1 value = input.createValue();

      while (input.next(key, value)) {
        while (!queue.offer(new KvPair<K1, V1>(key, value))
            && !future.isDone()) {
          // the queue is full
          Thread.yield();
        }
        // allocate new key & value instances as mapper is running in another
        // thread
        key = input.createKey();
        value = input.createValue();
      }

      while (!queue.offer(FINISH_TOKEN) && !future.isDone()) {
        // the queue is full
        Thread.yield();
      }

      // wait until the task is done or an exception was thrown
      try {
        future.get();
      } catch (ExecutionException ee) {
        throw new IOException(ee.getCause());
      } catch (InterruptedException inter) {
        Thread.currentThread().interrupt();
        // we don't need the result, so cancel the task
        future.cancel(true);
      }
    } finally {
      executorService.shutdownNow();
      mapper.close();
    }
  }
}

TestAsyncMapRunner.java

/**
 * 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.mapred.lib;

import java.io.DataOutputStream;
import java.io.IOException;
import java.util.Iterator;

import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;
import org.apache.hadoop.mapred.AsyncMapRunner;
import org.apache.hadoop.mapred.FileInputFormat;
import org.apache.hadoop.mapred.FileOutputFormat;
import org.apache.hadoop.mapred.HadoopTestCase;
import org.apache.hadoop.mapred.JobClient;
import org.apache.hadoop.mapred.JobConf;
import org.apache.hadoop.mapred.Mapper;
import org.apache.hadoop.mapred.OutputCollector;
import org.apache.hadoop.mapred.Reducer;
import org.apache.hadoop.mapred.Reporter;
import org.apache.hadoop.mapred.RunningJob;
import org.apache.hadoop.mapred.TextInputFormat;
import org.apache.hadoop.mapred.TextOutputFormat;

public class TestAsyncMapRunner extends HadoopTestCase {

  public TestAsyncMapRunner() throws IOException {
    super(HadoopTestCase.LOCAL_MR, HadoopTestCase.LOCAL_FS, 1, 1);
  }

  public void testOKRun() throws Exception {
    run(false, false);
  }

  public void testIOExRun() throws Exception {
    run(true, false);
  }
  public void testRuntimeExRun() throws Exception {
    run(false, true);
  }

  private void run(boolean ioEx, boolean rtEx) throws Exception {
    Path inDir = new Path("testing/mt/input");
    Path outDir = new Path("testing/mt/output");

    // Hack for local FS that does not have the concept of a 'mounting point'
    if (isLocalFS()) {
      String localPathRoot = System.getProperty("test.build.data", "/tmp")
              .replace(' ', '+');
      inDir = new Path(localPathRoot, inDir);
      outDir = new Path(localPathRoot, outDir);
    }


    JobConf conf = createJobConf();
    FileSystem fs = FileSystem.get(conf);

    fs.delete(outDir, true);
    if (!fs.mkdirs(inDir)) {
      throw new IOException("Mkdirs failed to create " + inDir.toString());
    }
    {
      DataOutputStream file = fs.create(new Path(inDir, "part-0"));
      file.writeBytes("a\nb\n\nc\nd\ne");
      file.close();
    }

    conf.setJobName("mt");
    conf.setInputFormat(TextInputFormat.class);

    conf.setOutputKeyClass(LongWritable.class);
    conf.setOutputValueClass(Text.class);

    conf.setMapOutputKeyClass(LongWritable.class);
    conf.setMapOutputValueClass(Text.class);

    conf.setOutputFormat(TextOutputFormat.class);
    conf.setOutputKeyClass(LongWritable.class);
    conf.setOutputValueClass(Text.class);

    conf.setMapperClass(IDMap.class);
    conf.setReducerClass(IDReduce.class);

    FileInputFormat.setInputPaths(conf, inDir);
    FileOutputFormat.setOutputPath(conf, outDir);

    conf.setMapRunnerClass(AsyncMapRunner.class);
    
    conf.setInt("mapred.map.asyncmaprunner.queue.size", 2);

    if (ioEx) {
      conf.setBoolean("async.ioException", true);
    }
    if (rtEx) {
      conf.setBoolean("async.runtimeException", true);
    }
    JobClient jc = new JobClient(conf);
    RunningJob job =jc.submitJob(conf);
    while (!job.isComplete()) {
      Thread.sleep(100);
    }

    if (job.isSuccessful()) {
      assertFalse(ioEx || rtEx);
    }
    else {
      assertTrue(ioEx || rtEx);
    }

  }

  public static class IDMap implements Mapper<LongWritable, Text,
                                              LongWritable, Text> {
    private boolean ioEx = false;
    private boolean rtEx = false;

    public void configure(JobConf job) {
      ioEx = job.getBoolean("async.ioException", false);
      rtEx = job.getBoolean("async.runtimeException", false);
    }

    public void map(LongWritable key, Text value,
                    OutputCollector<LongWritable, Text> output,
                    Reporter reporter)
            throws IOException {
      if (ioEx) {
        throw new IOException();
      }
      if (rtEx) {
        throw new RuntimeException();
      }
      output.collect(key, value);
      try {
        Thread.sleep(100);
      } catch (InterruptedException ex) {
        throw new RuntimeException(ex);
      }
    }


    public void close() throws IOException {
    }
  }

  public static class IDReduce implements Reducer<LongWritable, Text,
                                                  LongWritable, Text> {

    public void configure(JobConf job) {
    }

    public void reduce(LongWritable key, Iterator<Text> values,
                       OutputCollector<LongWritable, Text> output,
                       Reporter reporter)
            throws IOException {
      while (values.hasNext()) {
        output.collect(key, values.next());
      }
    }

    public void close() throws IOException {
    }
  }
}