HadoopSourceAnalyse --- Mapreduce ApplicationMaster Job start

Overview

MRAppMaster 初始化成功之后， Master 开始根据运行时的配置信息加Job并为之请求Container执行任务。

Start Job

   图 1-1

图 1-2

Code:

 amInfos = new LinkedList<AMInfo>();
    completedTasksFromPreviousRun = new HashMap<TaskId, TaskInfo>();
    processRecovery();

    // Current an AMInfo for the current AM generation.
    AMInfo amInfo =
        MRBuilderUtils.newAMInfo(appAttemptID, startTime, containerID, nmHost,
            nmPort, nmHttpPort);
    amInfos.add(amInfo);

    // /////////////////// Create the job itself.
    job = createJob(getConfig(), forcedState, shutDownMessage);

    // End of creating the job.

    // Send out an MR AM inited event for this AM and all previous AMs.
    for (AMInfo info : amInfos) {
      dispatcher.getEventHandler().handle(
          new JobHistoryEvent(job.getID(), new AMStartedEvent(info
              .getAppAttemptId(), info.getStartTime(), info.getContainerId(),
              info.getNodeManagerHost(), info.getNodeManagerPort(), info
                  .getNodeManagerHttpPort())));
    }

    // metrics system init is really init & start.
    // It's more test friendly to put it here.
    DefaultMetricsSystem.initialize("MRAppMaster");

    if (!errorHappenedShutDown) {
      // create a job event for job intialization
      JobEvent initJobEvent = new JobEvent(job.getID(), JobEventType.JOB_INIT);
      // Send init to the job (this does NOT trigger job execution)
      // This is a synchronous call, not an event through dispatcher. We want
      // job-init to be done completely here.
      jobEventDispatcher.handle(initJobEvent);


      // JobImpl's InitTransition is done (call above is synchronous), so the
      // "uber-decision" (MR-1220) has been made.  Query job and switch to
      // ubermode if appropriate (by registering different container-allocator
      // and container-launcher services/event-handlers).

      if (job.isUber()) {
        speculatorEventDispatcher.disableSpeculation();
        LOG.info("MRAppMaster uberizing job " + job.getID()
            + " in local container (\"uber-AM\") on node "
            + nmHost + ":" + nmPort + ".");
      } else {
        // send init to speculator only for non-uber jobs. 
        // This won't yet start as dispatcher isn't started yet.
        dispatcher.getEventHandler().handle(
            new SpeculatorEvent(job.getID(), clock.getTime()));
        LOG.info("MRAppMaster launching normal, non-uberized, multi-container "
            + "job " + job.getID() + ".");
      }
    }

    //start all the components
    super.start();

    // All components have started, start the job.
    startJobs();

在任务创建成功之后，MRAppMaster会注注册，JobFinishEventHandler 来处理 JobFinishEvent.Type;

现在任务已经提交到JobFSM，接下来，Job FSM 将继续执行请求的任务。