Java Code Examples for org.apache.tez.mapreduce.hadoop.MRHelpers#addMRInput()

The following examples show how to use org.apache.tez.mapreduce.hadoop.MRHelpers#addMRInput() . You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may check out the related API usage on the sidebar.
Example 1
Source File: YARNRunner.java    From incubator-tez with Apache License 2.0 4 votes vote down vote up 
private Vertex createVertexForStage(Configuration stageConf,
    Map<String, LocalResource> jobLocalResources,
    List<TaskLocationHint> locations, int stageNum, int totalStages)
    throws IOException {
  // stageNum starts from 0, goes till numStages - 1
  boolean isMap = false;
  if (stageNum == 0) {
    isMap = true;
  }

  int numTasks = isMap ? stageConf.getInt(MRJobConfig.NUM_MAPS, 0)
      : stageConf.getInt(MRJobConfig.NUM_REDUCES, 0);
  String processorName = isMap ? MapProcessor.class.getName()
      : ReduceProcessor.class.getName();
  String vertexName = null;
  if (isMap) {
    vertexName = MultiStageMRConfigUtil.getInitialMapVertexName();
  } else {
    if (stageNum == totalStages - 1) {
      vertexName = MultiStageMRConfigUtil.getFinalReduceVertexName();
    } else {
      vertexName = MultiStageMRConfigUtil
          .getIntermediateStageVertexName(stageNum);
    }
  }

  Resource taskResource = isMap ? MRHelpers.getMapResource(stageConf)
      : MRHelpers.getReduceResource(stageConf);
  
  stageConf.set(MRJobConfig.MROUTPUT_FILE_NAME_PREFIX, "part");
  
  byte[] vertexUserPayload = MRHelpers.createUserPayloadFromConf(stageConf);
  Vertex vertex = new Vertex(vertexName, new ProcessorDescriptor(processorName).
      setUserPayload(vertexUserPayload),
      numTasks, taskResource);
  if (isMap) {
    byte[] mapInputPayload = MRHelpers.createMRInputPayload(vertexUserPayload, null);
    MRHelpers.addMRInput(vertex, mapInputPayload, null);
  }
  // Map only jobs.
  if (stageNum == totalStages -1) {
    MRHelpers.addMROutputLegacy(vertex, vertexUserPayload);
  }

  Map<String, String> taskEnv = new HashMap<String, String>();
  setupMapReduceEnv(stageConf, taskEnv, isMap);

  Map<String, LocalResource> taskLocalResources =
      new TreeMap<String, LocalResource>();
  // PRECOMMIT Remove split localization for reduce tasks if it's being set
  // here
  taskLocalResources.putAll(jobLocalResources);

  String taskJavaOpts = isMap ? MRHelpers.getMapJavaOpts(stageConf)
      : MRHelpers.getReduceJavaOpts(stageConf);

  vertex.setTaskEnvironment(taskEnv)
      .setTaskLocalFiles(taskLocalResources)
      .setTaskLocationsHint(locations)
      .setTaskLaunchCmdOpts(taskJavaOpts);
  
  if (!isMap) {
    vertex.setVertexManagerPlugin(new VertexManagerPluginDescriptor(
        ShuffleVertexManager.class.getName()));
  }

  if (LOG.isDebugEnabled()) {
    LOG.debug("Adding vertex to DAG" + ", vertexName="
        + vertex.getName() + ", processor="
        + vertex.getProcessorDescriptor().getClassName() + ", parallelism="
        + vertex.getParallelism() + ", javaOpts=" + vertex.getTaskLaunchCmdOpts()
        + ", resources=" + vertex.getTaskResource()
    // TODO Add localResources and Environment
    );
  }

  return vertex;
}