Spark Streaming源码解读之Driver中的ReceiverTracker架构设计以及具体实现彻底研究

上节课讲到了Receiver是如何不断的接收数据的，并且接收到的数据的元数据会汇报给ReceiverTracker，下面我们看看ReceiverTracker具体的功能及实现。
一、 ReceiverTracker主要的功能：
在Executor上启动Receivers。
停止Receivers 。
更新Receiver接收数据的速率(也就是限流)
不断的等待Receivers的运行状态，只要Receivers停止运行，就重新启动Receiver。也就是Receiver的容错功能。
接受Receiver的注册。
借助ReceivedBlockTracker来管理Receiver接收数据的元数据。
汇报Receiver发送过来的错误信息


ReceiverTracker 管理了一个消息通讯体ReceiverTrackerEndpoint，用来与Receiver或者ReceiverTracker 进行消息通信。
在ReceiverTracker的start方法中，实例化了ReceiverTrackerEndpoint，并且在Executor上启动Receivers：
/** Start the endpoint and receiver execution thread. */
def start(): Unit = synchronized {
  if (isTrackerStarted) {
    throw new SparkException("ReceiverTracker already started")
  }
 
  if (!receiverInputStreams.isEmpty) {
    endpoint = ssc.env.rpcEnv.setupEndpoint(
      "ReceiverTracker", new ReceiverTrackerEndpoint(ssc.env.rpcEnv))
    if (!skipReceiverLaunch) launchReceivers()
    logInfo("ReceiverTracker started")
    trackerState = Started
  }
}


启动Receivr，其实是ReceiverTracker给ReceiverTrackerEndpoint发送了一个本地消息，ReceiverTrackerEndpoint将Receiver封装成RDD以job的方式提交给集群运行。
endpoint.send(StartAllReceivers(receivers))


这里的endpoint就是ReceiverTrackerEndpoint的引用。


Receiver启动后，会向ReceiverTracker注册，注册成功才算正式启动了。
override protected def onReceiverStart(): Boolean = {
  val msg = RegisterReceiver(
    streamId, receiver.getClass.getSimpleName, host, executorId, endpoint)
  trackerEndpoint.askWithRetry[Boolean](msg)
}


当Receiver端接收到数据，达到一定的条件需要将数据写入BlockManager，并且将数据的元数据汇报给ReceiverTracker：
/** Store block and report it to driver */
def pushAndReportBlock(
    receivedBlock: ReceivedBlock,
    metadataOption: Option[Any],
    blockIdOption: Option[StreamBlockId]
  ) {
  val blockId = blockIdOption.getOrElse(nextBlockId)
  val time = System.currentTimeMillis
  val blockStoreResult = receivedBlockHandler.storeBlock(blockId, receivedBlock)
  logDebug(s"Pushed block $blockId in ${(System.currentTimeMillis - time)} ms")
  val numRecords = blockStoreResult.numRecords
  val blockInfo = ReceivedBlockInfo(streamId, numRecords, metadataOption, blockStoreResult)
  trackerEndpoint.askWithRetry[Boolean](AddBlock(blockInfo))
  logDebug(s"Reported block $blockId")
}


当ReceiverTracker收到元数据后，会在线程池中启动一个线程来写数据：
case AddBlock(receivedBlockInfo) =>
  if (WriteAheadLogUtils.isBatchingEnabled(ssc.conf, isDriver = true)) {
    walBatchingThreadPool.execute(new Runnable {
      override def run(): Unit = Utils.tryLogNonFatalError {
        if (active) {
          context.reply(addBlock(receivedBlockInfo))
        } else {
          throw new IllegalStateException("ReceiverTracker RpcEndpoint shut down.")
        }
      }
    })
  } else {
    context.reply(addBlock(receivedBlockInfo))
  }
  
  数据的元数据是交由ReceivedBlockTracker管理的。
数据最终被写入到streamIdToUnallocatedBlockQueues中：一个流对应一个数据块信息的队列。


private type ReceivedBlockQueue = mutable.Queue[ReceivedBlockInfo]
 
private val streamIdToUnallocatedBlockQueues = new mutable.HashMap[Int, ReceivedBlockQueue]


每当Streaming 触发job时，会将队列中的数据分配成一个batch，并将数据写入timeToAllocatedBlocks数据结构。


private val timeToAllocatedBlocks = new mutable.HashMap[Time, AllocatedBlocks]
....
def allocateBlocksToBatch(batchTime: Time): Unit = synchronized {
  if (lastAllocatedBatchTime == null || batchTime > lastAllocatedBatchTime) {
    val streamIdToBlocks = streamIds.map { streamId =>
        (streamId, getReceivedBlockQueue(streamId).dequeueAll(x => true))
    }.toMap
    val allocatedBlocks = AllocatedBlocks(streamIdToBlocks)
    if (writeToLog(BatchAllocationEvent(batchTime, allocatedBlocks))) {
      timeToAllocatedBlocks.put(batchTime, allocatedBlocks)
      lastAllocatedBatchTime = batchTime
    } else {
      logInfo(s"Possibly processed batch $batchTime need to be processed again in WAL recovery")
    }
  } else {
    // This situation occurs when:
    // 1. WAL is ended with BatchAllocationEvent, but without BatchCleanupEvent,
    // possibly processed batch job or half-processed batch job need to be processed again,
    // so the batchTime will be equal to lastAllocatedBatchTime.
    // 2. Slow checkpointing makes recovered batch time older than WAL recovered
    // lastAllocatedBatchTime.
    // This situation will only occurs in recovery time.
    logInfo(s"Possibly processed batch $batchTime need to be processed again in WAL recovery")
  }
}


可见一个batch会包含多个流的数据。


每当Streaming 的一个job运行完毕后：
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) {
    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
    ))
    listenerBus.post(StreamingListenerBatchCompleted(jobSet.toBatchInfo))
  }
  ...
  
 JobScheduler会调用handleJobCompletion方法，最终会触发
 jobScheduler.receiverTracker.cleanupOldBlocksAndBatches(time - maxRememberDuration)
 
 这里的maxRememberDuration是DStream中每个时刻生成的RDD保留的最长时间。
 def cleanupOldBatches(cleanupThreshTime: Time, waitForCompletion: Boolean): Unit = synchronized {
  require(cleanupThreshTime.milliseconds < clock.getTimeMillis())
  val timesToCleanup = timeToAllocatedBlocks.keys.filter { _ < cleanupThreshTime }.toSeq
  logInfo("Deleting batches " + timesToCleanup)
  if (writeToLog(BatchCleanupEvent(timesToCleanup))) {
    timeToAllocatedBlocks --= timesToCleanup
    writeAheadLogOption.foreach(_.clean(cleanupThreshTime.milliseconds, waitForCompletion))
  } else {
    logWarning("Failed to acknowledge batch clean up in the Write Ahead Log.")
  }
}


而最后


listenerBus.post(StreamingListenerBatchCompleted(jobSet.toBatchInfo))


这个代码会调用
case batchCompleted: StreamingListenerBatchCompleted =>
  listener.onBatchCompleted(batchCompleted)
   
  ... 一路跟着下去...
   
  /**
 * A RateController that sends the new rate to receivers, via the receiver tracker.
 */
private[streaming] class ReceiverRateController(id: Int, estimator: RateEstimator)
    extends RateController(id, estimator) {
  override def publish(rate: Long): Unit =
    ssc.scheduler.receiverTracker.sendRateUpdate(id, rate)
}


/** Update a receiver's maximum ingestion rate */
def sendRateUpdate(streamUID: Int, newRate: Long): Unit = synchronized {
  if (isTrackerStarted) {
    endpoint.send(UpdateReceiverRateLimit(streamUID, newRate))
  }
}


case UpdateReceiverRateLimit(streamUID, newRate) =>
  for (info <- receiverTrackingInfos.get(streamUID); eP <- info.endpoint) {
    eP.send(UpdateRateLimit(newRate))
  }
  
发送调整速率的消息给Receiver，Receiver接到消息后，最终通过BlockGenerator来调整数据的写入的时间，而控制数据流的速率。


case UpdateRateLimit(eps) =>
  logInfo(s"Received a new rate limit: $eps.")
  registeredBlockGenerators.foreach { bg =>
    bg.updateRate(eps)
  }