spark DAGSchedulerEventProcessLoop 设计原理

父类EventLoop起了一个Thread，监听从LinkedBlockingDeque中获取event，然后用onReceive接收执行，DAGSchedulerEventProcessLoop类中onReceive方法调用了doOnReceive，具体判断事件的类别，并进行处理。

private[scheduler] class DAGSchedulerEventProcessLoop(dagScheduler: DAGScheduler)
  extends EventLoop[DAGSchedulerEvent]("dag-scheduler-event-loop") with Logging {

  private[this] val timer = dagScheduler.metricsSource.messageProcessingTimer

  /**
   * The main event loop of the DAG scheduler.
   */
  override def onReceive(event: DAGSchedulerEvent): Unit = {
    val timerContext = timer.time()
    try {
      doOnReceive(event)
    } finally {
      timerContext.stop()
    }
  }

  private def doOnReceive(event: DAGSchedulerEvent): Unit = event match {
// 处理job提交任务
    case JobSubmitted(jobId, rdd, func, partitions, callSite, listener, properties) =>
      dagScheduler.handleJobSubmitted(jobId, rdd, func, partitions, callSite, listener, properties)
  //处理map提交的stage任务
    case MapStageSubmitted(jobId, dependency, callSite, listener, properties) =>
      dagScheduler.handleMapStageSubmitted(jobId, dependency, callSite, listener, properties)
//处理map stage 取消的任务
    case StageCancelled(stageId, reason) =>
      dagScheduler.handleStageCancellation(stageId, reason)
//处理job 取消的任务
    case JobCancelled(jobId, reason) =>
      dagScheduler.handleJobCancellation(jobId, reason)
//处理job 组取消的任务
    case JobGroupCancelled(groupId) =>
      dagScheduler.handleJobGroupCancelled(groupId)
//处理所有job 取消的任务
    case AllJobsCancelled =>
      dagScheduler.doCancelAllJobs()
//处理executort完成分配的事件
    case ExecutorAdded(execId, host) =>
      dagScheduler.handleExecutorAdded(execId, host)
//Executor丢失
    case ExecutorLost(execId, reason) =>
      val filesLost = reason match {
        case SlaveLost(_, true) => true
        case _ => false
      }
      dagScheduler.handleExecutorLost(execId, filesLost)
//开始task
    case BeginEvent(task, taskInfo) =>
      dagScheduler.handleBeginEvent(task, taskInfo)
//获取task结果
    case GettingResultEvent(taskInfo) =>
      dagScheduler.handleGetTaskResult(taskInfo)
//事件完成
    case completion: CompletionEvent =>
      dagScheduler.handleTaskCompletion(completion)
//task失败
    case TaskSetFailed(taskSet, reason, exception) =>
      dagScheduler.handleTaskSetFailed(taskSet, reason, exception)
//重复提交
    case ResubmitFailedStages =>
      dagScheduler.resubmitFailedStages()
  }

  override def onError(e: Throwable): Unit = {
    logError("DAGSchedulerEventProcessLoop failed; shutting down SparkContext", e)
    try {
      dagScheduler.doCancelAllJobs()
    } catch {
      case t: Throwable => logError("DAGScheduler failed to cancel all jobs.", t)
    }
    dagScheduler.sc.stopInNewThread()
  }

  override def onStop(): Unit = {
    // Cancel any active jobs in postStop hook
    dagScheduler.cleanUpAfterSchedulerStop()
  }
}