关闭

第16课:Spark Streaming源码解读之数据清理内幕彻底解密

482人阅读 评论(0) 收藏 举报

 

本篇博客的主要目的是: 
1. 理清楚Spark Streaming中数据清理的流程

组织思路如下: 
a) 背景
b) 如何研究Spark Streaming数据清理?
c) 源码解析

一:背景 
Spark Streaming数据清理的工作无论是在实际开发中,还是自己动手实践中都是会面临的,Spark Streaming中Batch Durations中会不断的产生RDD,这样会不断的有内存对象生成,其中包含元数据和数据本身。由此Spark Streaming本身会有一套产生元数据以及数据的清理机制。

二:如何研究Spark Streaming数据清理?

1. 操作DStream的时候会产生元数据,所以要解决RDD的数据清理工作就一定要从DStream入手。因为DStream是RDD的模板,DStream之间有依赖关系。
DStream的操作产生了RDD,接收数据也靠DStream,数据的输入,数据的计算,输出整个生命周期都是由DStream构建的。由此,DStream负责RDD的整个生命周期。因此研究的入口的是DStream。

2. 基于Kafka数据来源,通过Direct的方式访问Kafka,DStream随着时间的进行,会不断的在自己的内存数据结构中维护一个 HashMap,HashMap维护的就是时间窗口,以及时间窗口下的RDD.按照Batch Duration来存储RDD以及删除RDD.

3. Spark Streaming本身是一直在运行的,在自己计算的时候会不断的产生RDD,例如每秒Batch Duration都会产生RDD,除此之外可能还有累加器,广播变量。由于不断的产生这些对象,因此Spark Streaming有自己的一套对象,元数据以及数据的清理机制。

4. Spark Streaming对RDD的管理就相当于JVM的GC.

三:源码解析 
generatedRDDs:安照Batch Duration的方式来存储RDD以及删除RDD。

// RDDs generated, marked as private[streaming] so that testsuites can access it

@transient

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

我们在实际开发中,可能手动缓存,即使不缓存的话,它在内存generatorRDD中也有对象,如何释放他们?不仅仅是RDD本身,也包括数据源(数据来源)和元数据(metada),因此释放RDD的时候这三方面都需要考虑。
释放跟时钟Click有关系,因为数据是周期性产生,所以肯定是周期性释放。 
因此下一步就需要找JobGenerator

1. RecurringTimer: 消息循环器将消息不断的发送给EventLoop

private val timer = new RecurringTimer(clock, ssc.graph.batchDuration.milliseconds,

  longTime => eventLoop.post(GenerateJobs(new Time(longTime))), "JobGenerator")

2.  eventLoop:onReceive接收到消息。

/** Start generation of jobs */

def start(): Unit = synchronized {

  if (eventLoop != null) return // generator has already been started

 

  // Call checkpointWriter here to initialize it before eventLoop uses it to avoid a deadlock.

  // See SPARK-10125

  checkpointWriter

 

  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)

    }

  }

3.  processEvent:中就会接收到ClearMetadata和ClearCheckpointData。

/** Processes all events */

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)

  }

}

4.  clearMetadata:清楚元数据信息。

/** Clear DStream metadata for the given `time`. */

private def clearMetadata(time: Time) {

  ssc.graph.clearMetadata(time)

 

  // If checkpointing is enabled, then checkpoint,

  // else mark batch to be fully processed

  if (shouldCheckpoint) {

    eventLoop.post(DoCheckpoint(time, clearCheckpointDataLater = true))

  } else {

    // If checkpointing is not enabled, then delete metadata information about

    // received blocks (block data not saved in any case). Otherwise, wait for

    // checkpointing of this batch to complete.

    val maxRememberDuration = graph.getMaxInputStreamRememberDuration()

    jobScheduler.receiverTracker.cleanupOldBlocksAndBatches(time - maxRememberDuration)

    jobScheduler.inputInfoTracker.cleanup(time - maxRememberDuration)

    markBatchFullyProcessed(time)

  }

}

5.  DStreamGraph:首先会清理outputDStream,其实就是forEachDStream

def clearMetadata(time: Time) {

  logDebug("Clearing metadata for time " + time)

  this.synchronized {

    outputStreams.foreach(_.clearMetadata(time))

  }

  logDebug("Cleared old metadata for time " + time)

}

6.  DStream.clearMetadata:除了清除RDD,也可以清除metadata元数据。如果想RDD跨Batch Duration的话可以设置rememberDuration时间. rememberDuration一般都是Batch Duration的倍数。

/**

 * Clear metadata that are older than `rememberDuration` of this DStream.

 * This is an internal method that should not be called directly. This default

 * implementation clears the old generated RDDs. Subclasses of DStream may override

 * this to clear their own metadata along with the generated RDDs.

 */

private[streaming] def clearMetadata(time: Time) {

  val unpersistData = ssc.conf.getBoolean("spark.streaming.unpersist", true)

// rememberDuration记忆周期 查看下RDD是否是oldRDD

  val oldRDDs = generatedRDDs.filter(_._1 <= (time - rememberDuration))

  logDebug("Clearing references to old RDDs: [" +

    oldRDDs.map(x => s"${x._1} -> ${x._2.id}").mkString(", ") + "]")

//从generatedRDDs中将key清理掉。

  generatedRDDs --= oldRDDs.keys

  if (unpersistData) {

    logDebug("Unpersisting old RDDs: " + oldRDDs.values.map(_.id).mkString(", "))

    oldRDDs.values.foreach { rdd =>

      rdd.unpersist(false)

      // Explicitly remove blocks of BlockRDD

      rdd match {

        case b: BlockRDD[_] =>

          logInfo("Removing blocks of RDD " + b + " of time " + time)

          b.removeBlocks() //清理掉RDD的数据

        case _ =>

      }

    }

  }

  logDebug("Cleared " + oldRDDs.size + " RDDs that were older than " +

    (time - rememberDuration) + ": " + oldRDDs.keys.mkString(", "))

//依赖的DStream也需要清理掉。

  dependencies.foreach(_.clearMetadata(time))

}

7.  在BlockRDD中,BlockManagerMaster根据blockId将Block删除。删除Block的操作是不可逆的。

/**

 * Remove the data blocks that this BlockRDD is made from. NOTE: This is an

 * irreversible operation, as the data in the blocks cannot be recovered back

 * once removed. Use it with caution.

 */

private[spark] def removeBlocks() {

  blockIds.foreach { blockId =>

    sparkContext.env.blockManager.master.removeBlock(blockId)

  }

  _isValid = false

}

回到上面JobGenerator中的processEvent 
1. clearCheckpoint:清除缓存数据。

/** Clear DStream checkpoint data for the given `time`. */

private def clearCheckpointData(time: Time) {

  ssc.graph.clearCheckpointData(time)

 

  // All the checkpoint information about which batches have been processed, etc have

  // been saved to checkpoints, so its safe to delete block metadata and data WAL files

  val maxRememberDuration = graph.getMaxInputStreamRememberDuration()

  jobScheduler.receiverTracker.cleanupOldBlocksAndBatches(time - maxRememberDuration)

  jobScheduler.inputInfoTracker.cleanup(time - maxRememberDuration)

  markBatchFullyProcessed(time)

}

2.  clearCheckpointData:

def clearCheckpointData(time: Time) {

  logInfo("Clearing checkpoint data for time " + time)

  this.synchronized {

    outputStreams.foreach(_.clearCheckpointData(time))

  }

  logInfo("Cleared checkpoint data for time " + time)

}

3.  ClearCheckpointData: 和清除元数据信息一样,还是清除DStream依赖的缓存数据。

private[streaming] def clearCheckpointData(time: Time) {

  logDebug("Clearing checkpoint data")

  checkpointData.cleanup(time)

  dependencies.foreach(_.clearCheckpointData(time))

  logDebug("Cleared checkpoint data")

}

4.  DStreamCheckpointData:清除缓存的数据

/**

 * Cleanup old checkpoint data. This gets called after a checkpoint of `time` has been

 * written to the checkpoint directory.

 */

def cleanup(time: Time) {

  // Get the time of the oldest checkpointed RDD that was written as part of the

  // checkpoint of `time`

  timeToOldestCheckpointFileTime.remove(time) match {

    case Some(lastCheckpointFileTime) =>

      // Find all the checkpointed RDDs (i.e. files) that are older than `lastCheckpointFileTime`

      // This is because checkpointed RDDs older than this are not going to be needed

      // even after master fails, as the checkpoint data of `time` does not refer to those files

      val filesToDelete = timeToCheckpointFile.filter(_._1 < lastCheckpointFileTime)

      logDebug("Files to delete:\n" + filesToDelete.mkString(","))

      filesToDelete.foreach {

        case (time, file) =>

          try {

            val path = new Path(file)

            if (fileSystem == null) {

              fileSystem = path.getFileSystem(dstream.ssc.sparkContext.hadoopConfiguration)

            }

            fileSystem.delete(path, true)

            timeToCheckpointFile -= time

            logInfo("Deleted checkpoint file '" + file + "' for time " + time)

          } catch {

            case e: Exception =>

              logWarning("Error deleting old checkpoint file '" + file + "' for time " + time, e)

              fileSystem = null

          }

      }

    case None =>

      logDebug("Nothing to delete")

  }

}

至此我们也知道了清理的过程,全流程如下: 

但是清理是什么时候被触发的? 
1. 在最终提交Job的时候,是交给JobHandler去执行的。

private class JobHandler(job: Job) extends Runnable with Logging {

    import JobScheduler._

 

    def run() {

      try {

        val formattedTime = UIUtils.formatBatchTime(

          job.time.milliseconds, ssc.graph.batchDuration.milliseconds, showYYYYMMSS = false)

        val batchUrl = s"/streaming/batch/?id=${job.time.milliseconds}"

        val batchLinkText = s"[output operation ${job.outputOpId}, batch time ${formattedTime}]"

 

        ssc.sc.setJobDescription(

          s"""Streaming job from <a href="$batchUrl">$batchLinkText</a>""")

        ssc.sc.setLocalProperty(BATCH_TIME_PROPERTY_KEY, job.time.milliseconds.toString)

        ssc.sc.setLocalProperty(OUTPUT_OP_ID_PROPERTY_KEY, job.outputOpId.toString)

 

        // We need to assign `eventLoop` to a temp variable. Otherwise, because

        // `JobScheduler.stop(false)` may set `eventLoop` to null when this method is running, then

        // it's possible that when `post` is called, `eventLoop` happens to null.

        var _eventLoop = eventLoop

        if (_eventLoop != null) {

          _eventLoop.post(JobStarted(job, clock.getTimeMillis()))

          // Disable checks for existing output directories in jobs launched by the streaming

          // scheduler, since we may need to write output to an existing directory during checkpoint

          // recovery; see SPARK-4835 for more details.

          PairRDDFunctions.disableOutputSpecValidation.withValue(true) {

            job.run()

          }

          _eventLoop = eventLoop

          if (_eventLoop != null) {

//当Job完成的时候,eventLoop会发消息初始化onReceive

            _eventLoop.post(JobCompleted(job, clock.getTimeMillis()))

          }

        } else {

          // JobScheduler has been stopped.

        }

      } finally {

        ssc.sc.setLocalProperty(JobScheduler.BATCH_TIME_PROPERTY_KEY, null)

        ssc.sc.setLocalProperty(JobScheduler.OUTPUT_OP_ID_PROPERTY_KEY, null)

      }

    }

  }

}

2.  OnReceive初始化接收到消息JobCompleted.

def start(): Unit = synchronized {

  if (eventLoop != null) return // scheduler has already been started

 

  logDebug("Starting JobScheduler")

  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()

3.  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)

  }

}

4.  调用JobGenerator的onBatchCompletion方法清楚元数据。

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))

  }

  job.result match {

    case Failure(e) =>

      reportError("Error running job " + job, e)

    case _ =>

  }

}

触发流程如下:

  • 资料来源于:DT_大数据梦工厂(Spark发行版本定制

  • DT大数据梦工厂微信公众号:DT_Spark 

  • 新浪微博:http://www.weibo.com/ilovepains

  • 王家林老师每晚20:00免费大数据实战

YY直播:68917580



 

0
0

查看评论
* 以上用户言论只代表其个人观点,不代表CSDN网站的观点或立场
    个人资料
    • 访问:5651次
    • 积分:203
    • 等级:
    • 排名:千里之外
    • 原创:15篇
    • 转载:1篇
    • 译文:0篇
    • 评论:0条
    文章存档