Example usage for org.apache.hadoop.mapreduce.server.tasktracker TTConfig TT_OUTOFBAND_HEARBEAT

Introduction

In this page you can find the example usage for org.apache.hadoop.mapreduce.server.tasktracker TTConfig TT_OUTOFBAND_HEARBEAT.

Prototype

String TT_OUTOFBAND_HEARBEAT

To view the source code for org.apache.hadoop.mapreduce.server.tasktracker TTConfig TT_OUTOFBAND_HEARBEAT.

Click Source Link

Usage

From source file:com.cloudera.oryx.computation.common.JobStep.java

License:Open Source License

/**
 * Creates a new {@link MRPipeline} instance that contains common configuration
 * settings./*from w w  w .  j a v a  2 s .  c  o  m*/
 *
 * @return a new {@link MRPipeline} instance, suitably configured
 */
protected final MRPipeline createBasicPipeline(Class<?> jarClass) throws IOException {
    Configuration conf = OryxConfiguration.get(getConf());

    conf.setBoolean(MRJobConfig.MAP_OUTPUT_COMPRESS, true);
    conf.setClass(MRJobConfig.MAP_OUTPUT_COMPRESS_CODEC, SnappyCodec.class, CompressionCodec.class);

    conf.setBoolean("mapred.output.compress", true);
    conf.set("mapred.output.compression.type", "BLOCK");
    conf.setClass("mapred.output.compression.codec", SnappyCodec.class, CompressionCodec.class);
    // Set old-style equivalents for Avro/Crunch's benefit
    conf.set("avro.output.codec", "snappy");

    conf.setBoolean(MRJobConfig.MAP_SPECULATIVE, true);
    conf.setBoolean(MRJobConfig.REDUCE_SPECULATIVE, true);
    conf.setBoolean(TTConfig.TT_OUTOFBAND_HEARBEAT, true);
    conf.setInt(MRJobConfig.JVM_NUMTASKS_TORUN, -1);

    //conf.setBoolean("crunch.disable.deep.copy", true);
    // Giving one mapper a lot of data can cause issues in some stages, so default to disable this
    conf.setBoolean("crunch.disable.combine.file", true);

    Config appConfig = ConfigUtils.getDefaultConfig();

    conf.set("crunch.tmp.dir", appConfig.getString("computation-layer.tmp-dir"));

    int mapMemoryMB = appConfig.getInt("computation-layer.mapper-memory-mb");
    log.info("Mapper memory: {}", mapMemoryMB);
    int mapHeapMB = (int) (mapMemoryMB / 1.3); // Matches Hadoop's default
    log.info("Mappers have {}MB heap and can access {}MB RAM", mapHeapMB, mapMemoryMB);
    if (conf.get(MRJobConfig.MAP_JAVA_OPTS) != null) {
        log.info("Overriding previous setting of {}, which was '{}'", MRJobConfig.MAP_JAVA_OPTS,
                conf.get(MRJobConfig.MAP_JAVA_OPTS));
    }
    conf.set(MRJobConfig.MAP_JAVA_OPTS,
            "-Xmx" + mapHeapMB + "m -XX:+UseCompressedOops -XX:+UseParallelGC -XX:+UseParallelOldGC");
    log.info("Set {} to '{}'", MRJobConfig.MAP_JAVA_OPTS, conf.get(MRJobConfig.MAP_JAVA_OPTS));
    // See comment below on CM
    conf.setInt("mapreduce.map.java.opts.max.heap", mapHeapMB);

    int reduceMemoryMB = appConfig.getInt("computation-layer.reducer-memory-mb");
    log.info("Reducer memory: {}", reduceMemoryMB);
    if (isHighMemoryStep()) {
        reduceMemoryMB *= appConfig.getInt("computation-layer.worker-high-memory-factor");
        log.info("Increasing {} to {} for high-memory step", MRJobConfig.REDUCE_MEMORY_MB, reduceMemoryMB);
    }
    conf.setInt(MRJobConfig.REDUCE_MEMORY_MB, reduceMemoryMB);

    int reduceHeapMB = (int) (reduceMemoryMB / 1.3); // Matches Hadoop's default
    log.info("Reducers have {}MB heap and can access {}MB RAM", reduceHeapMB, reduceMemoryMB);
    if (conf.get(MRJobConfig.REDUCE_JAVA_OPTS) != null) {
        log.info("Overriding previous setting of {}, which was '{}'", MRJobConfig.REDUCE_JAVA_OPTS,
                conf.get(MRJobConfig.REDUCE_JAVA_OPTS));
    }
    conf.set(MRJobConfig.REDUCE_JAVA_OPTS,
            "-Xmx" + reduceHeapMB + "m -XX:+UseCompressedOops -XX:+UseParallelGC -XX:+UseParallelOldGC");
    log.info("Set {} to '{}'", MRJobConfig.REDUCE_JAVA_OPTS, conf.get(MRJobConfig.REDUCE_JAVA_OPTS));
    // I see this in CM but not in Hadoop docs; probably won't hurt as it's supposed to result in
    // -Xmx appended to opts above, which is at worst redundant
    conf.setInt("mapreduce.reduce.java.opts.max.heap", reduceHeapMB);

    conf.setInt("yarn.scheduler.capacity.minimum-allocation-mb", 128);
    conf.setInt("yarn.app.mapreduce.am.resource.mb", 384);

    // Pass total config state
    conf.set(CONFIG_SERIALIZATION_KEY, ConfigUtils.getDefaultConfig().root().render());

    // Make sure to set any args to conf above this line!

    setConf(conf);

    Job job = Job.getInstance(conf);

    // Basic File IO settings
    FileInputFormat.setMaxInputSplitSize(job, 1L << 28); // ~268MB
    SequenceFileOutputFormat.setOutputCompressionType(job, SequenceFile.CompressionType.BLOCK);
    FileOutputFormat.setCompressOutput(job, true);
    FileOutputFormat.setOutputCompressorClass(job, SnappyCodec.class);

    log.info("Created pipeline configuration {}", job.getConfiguration());

    return new MRPipeline(jarClass, getCustomJobName(), job.getConfiguration());
}

---