本文主要通过源码来了解SparkStreaming程序从任务生成到任务完成整个执行流程以及中间伴随的checkpoint操作

注：下面源码只贴出跟分析内容有关的代码，其他省略

1 结论先行

SparkStreaming的启动、任务生成、任务结束、Checkpoint操作流程如下：

• SparkStreamingContext.start() 启动 JobScheduler
• JobScheduler的启动操作
  • JobScheduler 创建了 EventLoop[JobSchedulerEvent] 来处理 JobStarted 和 JobCompleted 事件
  • 启动 JobGenerator
• JobGenerator 的启动操作
  • JobGenerator 创建了 EventLoop[JobGeneratorEvent] 来处理 GenerateJobs、ClearMetaData、DoCheckPoint和ClearCheckpointData 事件
  • 创建RecurringTimer周期性地发送 GenerateJobs 事件用于任务的周期性创建和执行
• JobGenerator的任务生成工作
  • 调用 geneateJobs() 来生成 JobSet 并通过 JobScheduler.submitJobset 进行任务的提交
    • submitJobset 将 Job 作为参数传入 JobHandler ，并将 JobHandler 丢进线程池 JobExecutor 中执行
  • 发送 DoCheckPoint 事件进行元数据的 checkpoint
• 任务完成
  • JobHandler 中任务完成之后会发送 JobCompleted 事件，JobScheduler.EventLoop 接收到该事件后会进行处理，并且判断 JobSet 全部完成之后，发送 ClearMetaData 事件，进行数据的 checkpoint 或者删除

2 具体分析

应用程序入口：

val sparkConf = new SparkConf().setAppName("SparkStreaming")
val sc = new SparkContext(sparkConf)
val ssc = new StreamingContext(sc, Seconds(batchDuration.toLong))
ssc.start()
ssc.awaitTermination()

一旦ssc.start()调用之后，程序便真正开始运行

第一步：
SparkStreamingContext.start()进行如下主要工作：

• 调用JobScheduler.start()
• 发送StreamingListenerStreamingStarted消息

SparkStreamingContext.start()

def start(): Unit = synchronized {
    state match {
      case INITIALIZED =>
        StreamingContext.ACTIVATION_LOCK.synchronized {
          StreamingContext.assertNoOtherContextIsActive()
          try｛
              ...
              scheduler.start()
            }
            state = StreamingContextState.ACTIVE
            scheduler.listenerBus.post(
              StreamingListenerStreamingStarted(System.currentTimeMillis()))
          } catch {
            ...
          }
          StreamingContext.setActiveContext(this)
        }
        ...
      case ACTIVE =>
        logWarning("StreamingContext has already been started")
      case STOPPED =>
        throw new IllegalStateException("StreamingContext has already been stopped")
    }
  }

第二步：
调用JobScheduler.start()执行以下主要操作：

• 创建EventLoop用于处理接收到的JobSchedulerEvent，processEvent就是实际的处理逻辑
• 调用jobGenerator.start()

JobScheduler.start()：

def start(): Unit = synchronized {
    if (eventLoop != null) return // scheduler has already been started

    logDebug("Starting JobScheduler")
    //创建一个Event监听器并启动
    eventLoop = new EventLoop[JobSchedulerEvent]("JobScheduler") {
      override protected def onReceive(event: JobSchedulerEvent): Unit = processEvent(event)

      override protected def onError(e: Throwable): Unit = reportError("Error in job scheduler", e)
    }
    eventLoop.start()
    ...
    //启动JobGenerator
    jobGenerator.start()
    ...
  }

第三步：
JobGenerator.start()执行以下主要操作：

• 创建EventLoop[JobGeneratorEvent]用于处理JobGeneratorEvent事件
• 开始执行job的生成工作
  • 创建一个timer周期地执行eventLoop.post(GenerateJobs(new Time(longTime)))
  • JobGenerator.start()中的EventLoop收到GenerateJobs事件后，去执行generateJobs(time)
  • generateJobs()中生成JobSet并调用jobScheduler.submitJobSet()进行提交，然后发送一个DoCheckpointEvent进行checkpoint

JobGenerator.start()

def start(): Unit = synchronized {
    if (eventLoop != null) return // generator has already been started
    //创建checkpointWriter用于将checkpoint信息持久化
    checkpointWriter
    //创建了Event监听器，用于监听JobGeneratorEvent并处理
    eventLoop = new EventLoop[JobGeneratorEvent]("JobGenerator") {
      override protected def onReceive(event: JobGeneratorEvent): Unit = processEvent(event)

      override protected def onError(e: Throwable): Unit = {
        jobScheduler.reportError("Error in job generator", e)
      }
    }
    eventLoop.start()

    if (ssc.isCheckpointPresent) {
      //从checkpoint中恢复
      restart()
    } else {
      //首次创建
      startFirstTime()
    }
}

首次启动会调用startFirstTime()，在该方法中主要是调用已经初始化好的RecurringTimer.start()进行周期性的发送GenerateJobs事件，这个周期是ssc.graph.batchDuration.milliseconds也就是你所设置的batchTime，JobGenerate.start()中所创建好的EventLoop收到GenerateJobs事件，就会执行processEvent(),从而执行generateJobs(time)来进行Job的生成工作

  private val timer = new RecurringTimer(clock, ssc.graph.batchDuration.milliseconds,
    longTime => eventLoop.post(GenerateJobs(new Time(longTime))), "JobGenerator")

  private def startFirstTime() {
    val startTime = new Time(timer.getStartTime())
    graph.start(startTime - graph.batchDuration)
    timer.start(startTime.milliseconds)
    logInfo("Started JobGenerator at " + startTime)
  }

  private def processEvent(event: JobGeneratorEvent) {
    logDebug("Got event " + event)
    event match {
      case GenerateJobs(time) => generateJobs(time)
      case ClearMetadata(time) => clearMetadata(time)
      case DoCheckpoint(time, clearCheckpointDataLater) =>
        doCheckpoint(time, clearCheckpointDataLater)
      case ClearCheckpointData(time) => clearCheckpointData(time)
    }
  }

generateJobs的主要工作：

• 生成JobSet并调用jobScheduler.submitJobSet()进行提交
• 发送一个DoCheckpointEvent进行checkpoint

  private def generateJobs(time: Time) {
    ssc.sparkContext.setLocalProperty(RDD.CHECKPOINT_ALL_MARKED_ANCESTORS, "true")
    Try {
      jobScheduler.receiverTracker.allocateBlocksToBatch(time) // allocate received blocks to batch
      graph.generateJobs(time) // generate jobs using allocated block
    } match {
      case Success(jobs) =>
        val streamIdToInputInfos = jobScheduler.inputInfoTracker.getInfo(time)
        jobScheduler.submitJobSet(JobSet(time, jobs, streamIdToInputInfos))
      case Failure(e) =>
        jobScheduler.reportError("Error generating jobs for time " + time, e)
        PythonDStream.stopStreamingContextIfPythonProcessIsDead(e)
    }
    eventLoop.post(DoCheckpoint(time, clearCheckpointDataLater = false))
  }

第一个操作：jobScheduler.submitJobSet()中的主要操作是遍历jobSet中的job，并将其作为参数传入JobHandler对象中，并将JobHandler丢到jobExecutor中去执行。JobHandler是实现了Runnable，它的run()主要做了以下三件事

• 发送JobStarted事件
• 执行Job.run()
• 发送JobCompleted事件

def submitJobSet(jobSet: JobSet) {
    if (jobSet.jobs.isEmpty) {
      logInfo("No jobs added for time " + jobSet.time)
    } else {
      listenerBus.post(StreamingListenerBatchSubmitted(jobSet.toBatchInfo))
      jobSets.put(jobSet.time, jobSet)
      jobSet.jobs.foreach(job => jobExecutor.execute(new JobHandler(job)))
      logInfo("Added jobs for time " + jobSet.time)
    }
}

private class JobHandler(job: Job) extends Runnable with Logging {
    import JobScheduler._
    def run() {
      try {
        var _eventLoop = eventLoop
        if (_eventLoop != null) {
          _eventLoop.post(JobStarted(job, clock.getTimeMillis()))//发送JobStarted事件
          SparkHadoopWriterUtils.disableOutputSpecValidation.withValue(true) {
            job.run()
          }
          _eventLoop = eventLoop
          if (_eventLoop != null) {
            _eventLoop.post(JobCompleted(job, clock.getTimeMillis()))//发送JobCompleted事件
          }
        } else {
        }
      } finally {
        ssc.sparkContext.setLocalProperties(oldProps)
      }
    }
  }

第二个操作：发送DoCheckpoint事件

JobScheduler.start()中创建的EventLoop的核心内容是processEvent(event)方法，Event的类型有三种，分别是JobStarted、JobCompleted和ErrorReported，EventLoop收到DoCheckpoint事件后会执行doCheckpoint()：

  //JobGenerator.processEvent()
  private def processEvent(event: JobGeneratorEvent) {
    logDebug("Got event " + event)
    event match {
      ...
      case DoCheckpoint(time, clearCheckpointDataLater) =>
        doCheckpoint(time, clearCheckpointDataLater)
      ...
    }
  }

doCheckpoint()调用graph.updateCheckpointData进行checkpoint信息的更新，调用checkpointWriter.write对checkpoint信息进行持久化

  private def doCheckpoint(time: Time, clearCheckpointDataLater: Boolean) {
    if (shouldCheckpoint && (time - graph.zeroTime).isMultipleOf(ssc.checkpointDuration)) {
      logInfo("Checkpointing graph for time " + time)
      //将新的checkpoint写到
      ssc.graph.updateCheckpointData(time)
      //将checkpoint写到文件系统中
      checkpointWriter.write(new Checkpoint(ssc, time), clearCheckpointDataLater)
    } else if (clearCheckpointDataLater) {
      markBatchFullyProcessed(time)
    }
  }

checkpoint的update中主要是调用DStreamGraph.updateCheckpointData：

def updateCheckpointData(time: Time) {
    logInfo("Updating checkpoint data for time " + time)
    this.synchronized {
      outputStreams.foreach(_.updateCheckpointData(time))
    }
    logInfo("Updated checkpoint data for time " + time)
  }

outputStreams.foreach(_.updateCheckpointData(time))则是调用了DStream中的updateCheckpointData方法，而该方法主要是调用checkpointData.update(currentTime)来进行更新，并且调用该DStream所依赖的DStream进行更新。

private[streaming] var generatedRDDs = new HashMap[Time, RDD[T]]()

private[streaming] def updateCheckpointData(currentTime: Time) {
    logDebug(s"Updating checkpoint data for time $currentTime")
    checkpointData.update(currentTime)
    dependencies.foreach(_.updateCheckpointData(currentTime))
    logDebug(s"Updated checkpoint data for time $currentTime: $checkpointData")
  }

我们接下来来看看checkpointData.update(currentTime)：我们可以在DStream中看到以下的实现：

private[streaming] val checkpointData = new DStreamCheckpointData(this)

我们接着找到了：DStreamCheckpointData.update，DStreamCheckpointData有其他子类用于自定义保存的内容和逻辑

  //key为指定时间，value为checkpoint file内容
  @transient private var timeToCheckpointFile = new HashMap[Time, String]
  // key为batchtime，value该batch中最先checkpointed RDD的time
  @transient private var timeToOldestCheckpointFileTime = new HashMap[Time, Time]
  protected[streaming] def currentCheckpointFiles = data.asInstanceOf[HashMap[Time, String]]

def update(time: Time) {
    //从dsteam中获得要checkpoint的RDDs,generatedRDDs就是一个HashMap[Time, RDD[T]]
    val checkpointFiles = dstream.generatedRDDs.filter(_._2.getCheckpointFile.isDefined)
                                       .map(x => (x._1, x._2.getCheckpointFile.get))
    logDebug("Current checkpoint files:\n" + checkpointFiles.toSeq.mkString("\n"))

    // checkpoint文件添加到最后要进行序列化的HashMap中
    if (!checkpointFiles.isEmpty) {
      currentCheckpointFiles.clear()
      currentCheckpointFiles ++= checkpointFiles
      //更新checkpointfile
      timeToCheckpointFile ++= currentCheckpointFiles
      // key为传入的time，value为最先进行checkpoint的rdd的time
      timeToOldestCheckpointFileTime(time) = currentCheckpointFiles.keys.min(Time.ordering)
    }
  }

第四步：任务完成
在上面generateJobs中所调用的jobScheduler.submitJobSet()中提到每个Job都会作为参数传入JobHandler,而JobHandler会丢到JobExecutor中去执行，而JobHandler的主要工作是发送JobStarted事件，执行完任务后会发送JobCompleted事件，而JobScheduler.EventLoop收到事件后会执行handleJobCompletion方法

 //JobScheduler.processEvent()
 private def processEvent(event: JobSchedulerEvent) {
    try {
      event match {
        case JobStarted(job, startTime) => handleJobStart(job, startTime)
        case JobCompleted(job, completedTime) => handleJobCompletion(job, completedTime)
        case ErrorReported(m, e) => handleError(m, e)
      }
    } catch {
      case e: Throwable =>
        reportError("Error in job scheduler", e)
    }
  }

handleJobCompletion方法会调用jobSet.handleJobCompletion(job)，并且在最后会判断jobSet是否已经全部完成，如果是的话会执行jobGenerator.onBatchCompletion(jobSet.time)

private def handleJobCompletion(job: Job, completedTime: Long) {
    val jobSet = jobSets.get(job.time)
    jobSet.handleJobCompletion(job)
    job.setEndTime(completedTime)
    listenerBus.post(StreamingListenerOutputOperationCompleted(job.toOutputOperationInfo))
    logInfo("Finished job " + job.id + " from job set of time " + jobSet.time)
    if (jobSet.hasCompleted) {
      listenerBus.post(StreamingListenerBatchCompleted(jobSet.toBatchInfo))
    }
    job.result match {
      case Failure(e) =>
        reportError("Error running job " + job, e)
      case _ => //如果所有事件完成则会执行以下操作
        if (jobSet.hasCompleted) {
          jobSets.remove(jobSet.time)
          jobGenerator.onBatchCompletion(jobSet.time)
          logInfo("Total delay: %.3f s for time %s (execution: %.3f s)".format(
            jobSet.totalDelay / 1000.0, jobSet.time.toString,
            jobSet.processingDelay / 1000.0
          ))
        }
    }
  }

此时到JobGenerator中找到onBatchCompletion():

def onBatchCompletion(time: Time) {
    eventLoop.post(ClearMetadata(time))
}

JobGenerator.processEvent()执行clearMetadata(time)

private def processEvent(event: JobGeneratorEvent) {
    logDebug("Got event " + event)
    event match {
      case GenerateJobs(time) => generateJobs(time)
      case ClearMetadata(time) => clearMetadata(time)
      case DoCheckpoint(time, clearCheckpointDataLater) =>
        doCheckpoint(time, clearCheckpointDataLater)
      case ClearCheckpointData(time) => clearCheckpointData(time)
    }
}

clearMetadata()对原数据进行checkpoint或者删除

private def clearMetadata(time: Time) {
    ssc.graph.clearMetadata(time)

    // If checkpointing is enabled, then checkpoint,
    // else mark batch to be fully processed
    if (shouldCheckpoint) {
      //如果需要进行checkpoint，发送DoCheckpoint事件 
      eventLoop.post(DoCheckpoint(time, clearCheckpointDataLater = true))
    } else {
      //将meta数据进行删除
    }
}

附：RecurringTimer和EventLoop的源码，并作简单的注释

RecurringTimer的代码如下：

private[streaming]
class RecurringTimer(clock: Clock, period: Long, callback: (Long) => Unit, name: String)
  extends Logging {
  //创建一个thread，用来执行loop
  private val thread = new Thread("RecurringTimer - " + name) {
    setDaemon(true)
    override def run() { loop }
  }

  @volatile private var prevTime = -1L
  @volatile private var nextTime = -1L
  @volatile private var stopped = false

  def getStartTime(): Long = {
    (math.floor(clock.getTimeMillis().toDouble / period) + 1).toLong * period
  }

  def getRestartTime(originalStartTime: Long): Long = {
    val gap = clock.getTimeMillis() - originalStartTime
    (math.floor(gap.toDouble / period).toLong + 1) * period + originalStartTime
  }

  //start方法中主要是启动thread，用于执行thread中的loop
  def start(startTime: Long): Long = synchronized {
    nextTime = startTime
    thread.start()
    logInfo("Started timer for " + name + " at time " + nextTime)
    nextTime
  }

  def start(): Long = {
    start(getStartTime())
  }

  def stop(interruptTimer: Boolean): Long = synchronized {
    if (!stopped) {
      stopped = true
      if (interruptTimer) {
        thread.interrupt()
      }
      thread.join()
      logInfo("Stopped timer for " + name + " after time " + prevTime)
    }
    prevTime
  }

  private def triggerActionForNextInterval(): Unit = {
    clock.waitTillTime(nextTime)
    callback(nextTime)
    prevTime = nextTime
    nextTime += period
    logDebug("Callback for " + name + " called at time " + prevTime)
  }

  //周期性地执行callback的内容，也就是 
  private def loop() {
    try {
      while (!stopped) {
        triggerActionForNextInterval()
      }
      triggerActionForNextInterval()
    } catch {
      case e: InterruptedException =>
    }
  }
}

EventLoop的源码：主要功能就是创建一个线程在后台判断是否Event进来，有的话则进行相应的处理

private[spark] abstract class EventLoop[E](name: String) extends Logging {

  private val eventQueue: BlockingQueue[E] = new LinkedBlockingDeque[E]()

  private val stopped = new AtomicBoolean(false)

  private val eventThread = new Thread(name) {
    setDaemon(true)

    override def run(): Unit = {
      try {
        while (!stopped.get) {
          val event = eventQueue.take()
          try {
            onReceive(event)
          } catch {
            case NonFatal(e) =>
              try {
                onError(e)
              } catch {
                case NonFatal(e) => logError("Unexpected error in " + name, e)
              }
          }
        }
      } catch {
        case ie: InterruptedException => // exit even if eventQueue is not empty
        case NonFatal(e) => logError("Unexpected error in " + name, e)
      }
    }

  }

  def start(): Unit = {
    if (stopped.get) {
      throw new IllegalStateException(name + " has already been stopped")
    }
    // Call onStart before starting the event thread to make sure it happens before onReceive
    onStart()
    eventThread.start()
  }

  def stop(): Unit = {
    if (stopped.compareAndSet(false, true)) {
      eventThread.interrupt()
      var onStopCalled = false
      try {
        eventThread.join()
        // Call onStop after the event thread exits to make sure onReceive happens before onStop
        onStopCalled = true
        onStop()
      } catch {
        case ie: InterruptedException =>
          Thread.currentThread().interrupt()
          if (!onStopCalled) {
            // ie is thrown from `eventThread.join()`. Otherwise, we should not call `onStop` since
            // it's already called.
            onStop()
          }
      }
    } else {
      // Keep quiet to allow calling `stop` multiple times.
    }
  }

  //将event放进eventQueue中，在eventThread进行相应的处理
  def post(event: E): Unit = {
    eventQueue.put(event)
  }

  //返回eventThread是否存活
  def isActive: Boolean = eventThread.isAlive

  //eventThread启动前调用
  protected def onStart(): Unit = {}

  //在eventThred结束后调用
  protected def onStop(): Unit = {}

  //实现类实现Event的处理逻辑
  protected def onReceive(event: E): Unit

  //出错时的处理逻辑
  protected def onError(e: Throwable): Unit

}