《深入理解SPARK：核心思想与源码分析》——SparkContext的初始化（中）

  protected val sparkListeners = new ArrayBuffer[SparkListener]
    with mutable.SynchronizedBuffer[SparkListener]

  def addListener(listener: SparkListener) {
    sparkListeners += listener
  }

  def postToAll(event: SparkListenerEvent) {
    event match {
      case stageSubmitted: SparkListenerStageSubmitted =>
        foreachListener(_.onStageSubmitted(stageSubmitted))
      case stageCompleted: SparkListenerStageCompleted =>
        foreachListener(_.onStageCompleted(stageCompleted))
      case jobStart: SparkListenerJobStart =>
        foreachListener(_.onJobStart(jobStart))
      case jobEnd: SparkListenerJobEnd =>
        foreachListener(_.onJobEnd(jobEnd))
      case taskStart: SparkListenerTaskStart =>
        foreachListener(_.onTaskStart(taskStart))
      case taskGettingResult: SparkListenerTaskGettingResult =>
        foreachListener(_.onTaskGettingResult(taskGettingResult))
      case taskEnd: SparkListenerTaskEnd =>
        foreachListener(_.onTaskEnd(taskEnd))
      case environmentUpdate: SparkListenerEnvironmentUpdate =>
        foreachListener(_.onEnvironmentUpdate(environmentUpdate))
      case blockManagerAdded: SparkListenerBlockManagerAdded =>
        foreachListener(_.onBlockManagerAdded(blockManagerAdded))
      case blockManagerRemoved: SparkListenerBlockManagerRemoved =>
        foreachListener(_.onBlockManagerRemoved(blockManagerRemoved))
      case unpersistRDD: SparkListenerUnpersistRDD =>
        foreachListener(_.onUnpersistRDD(unpersistRDD))
      case applicationStart: SparkListenerApplicationStart =>
        foreachListener(_.onApplicationStart(applicationStart))
      case applicationEnd: SparkListenerApplicationEnd =>
        foreachListener(_.onApplicationEnd(applicationEnd))
      case metricsUpdate: SparkListenerExecutorMetricsUpdate =>
        foreachListener(_.onExecutorMetricsUpdate(metricsUpdate))
      case SparkListenerShutdown =>
    }
  }

  private def foreachListener(f: SparkListener => Unit): Unit = {
    sparkListeners.foreach { listener =>
      try {
        f(listener)
      } catch {
        case e: Exception =>
          logError(s"Listener ${Utils.getFormattedClassName(listener)} threw an exception", e)
      }
    }
  }

  private[spark] val jobProgressListener = new JobProgressListener(conf)
  listenerBus.addListener(jobProgressListener)

  val statusTracker = new SparkStatusTracker(this)

  private[spark] val ui: Option[SparkUI] =
    if (conf.getBoolean("spark.ui.enabled", true)) {
      Some(SparkUI.createLiveUI(this, conf, listenerBus, jobProgressListener,
        env.securityManager,appName))
    } else {
      None
    }

  ui.foreach(_.bind())

  def createLiveUI(
      sc: SparkContext,
      conf: SparkConf,
      listenerBus: SparkListenerBus,
      jobProgressListener: JobProgressListener,
      securityManager: SecurityManager,
      appName: String): SparkUI =  {
    create(Some(sc), conf, listenerBus, securityManager, appName,
      jobProgressListener = Some(jobProgressListener))
  }

  private def create(
      sc: Option[SparkContext],
      conf: SparkConf,
      listenerBus: SparkListenerBus,
      securityManager: SecurityManager,
      appName: String,
      basePath: String = "",
      jobProgressListener: Option[JobProgressListener] = None): SparkUI = {

    val _jobProgressListener: JobProgressListener = jobProgressListener.getOrElse {
      val listener = new JobProgressListener(conf)
      listenerBus.addListener(listener)
      listener
    }

    val environmentListener = new EnvironmentListener
    val storageStatusListener = new StorageStatusListener
    val executorsListener = new ExecutorsListener(storageStatusListener)
    val storageListener = new StorageListener(storageStatusListener)

    listenerBus.addListener(environmentListener)
    listenerBus.addListener(storageStatusListener)
    listenerBus.addListener(executorsListener)
    listenerBus.addListener(storageListener)

    new SparkUI(sc, conf, securityManager, environmentListener, storageStatusListener,
      executorsListener, _jobProgressListener, storageListener, appName, basePath)
  }

private[ui] class JobsTab(parent: SparkUI) extends SparkUITab(parent, "jobs") {
  val sc = parent.sc
  val killEnabled = parent.killEnabled
  def isFairScheduler = listener.schedulingMode.exists(_ == SchedulingMode.FAIR)
  val listener = parent.jobProgressListener

  attachPage(new AllJobsPage(this))
  attachPage(new JobPage(this))
} 

def render(request: HttpServletRequest): Seq[Node] = {
    listener.synchronized {
      val activeJobs = listener.activeJobs.values.toSeq
      val completedJobs = listener.completedJobs.reverse.toSeq
      val failedJobs = listener.failedJobs.reverse.toSeq
      val now = System.currentTimeMillis

      val activeJobsTable =
        jobsTable(activeJobs.sortBy(_.startTime.getOrElse(-1L)).reverse)
      val completedJobsTable =
        jobsTable(completedJobs.sortBy(_.endTime.getOrElse(-1L)).reverse)
      val failedJobsTable =
        jobsTable(failedJobs.sortBy(_.endTime.getOrElse(-1L)).reverse)

      val summary: NodeSeq =
        <div>
          <ul class="unstyled">
            {if (startTime.isDefined) {
              // Total duration is not meaningful unless the UI is live
              <li>
                <strong>Total Duration: </strong>
                {UIUtils.formatDuration(now - startTime.get)}
              </li>
            }}
            <li>
              <strong>Scheduling Mode: </strong>
              {listener.schedulingMode.map(_.toString).getOrElse("Unknown")}
            </li>
            <li>
              <a href="#active"><strong>Active Jobs:</strong></a>
              {activeJobs.size}
            </li>
            <li>
              <a href="#completed"><strong>Completed Jobs:</strong></a>
              {completedJobs.size}
            </li>
            <li>
              <a href="#failed"><strong>Failed Jobs:</strong></a>
              {failedJobs.size}
            </li>
          </ul>
        </div> 

private def jobsTable(jobs: Seq[JobUIData]): Seq[Node] = {
    val someJobHasJobGroup = jobs.exists(_.jobGroup.isDefined)

    val columns: Seq[Node] = {
      <th>{if (someJobHasJobGroup) "Job Id (Job Group)" else "Job Id"}</th>
      <th>Description</th>
      <th>Submitted</th>
      <th>Duration</th>
      <th class="sorttable_nosort">Stages: Succeeded/Total</th>
      <th class="sorttable_nosort">Tasks (for all stages): Succeeded/Total</th>
    }

    <table class="table table-bordered table-striped table-condensed sortable">
      <thead>{columns}</thead>
      <tbody>
        {jobs.map(makeRow)}
      </tbody>
    </table>
  }

def makeRow(job: JobUIData): Seq[Node] = {
      val lastStageInfo = Option(job.stageIds)
        .filter(_.nonEmpty)
        .flatMap { ids => listener.stageIdToInfo.get(ids.max) }
      val lastStageData = lastStageInfo.flatMap { s =>
        listener.stageIdToData.get((s.stageId, s.attemptId))
      }
      val isComplete = job.status == JobExecutionStatus.SUCCEEDED
      val lastStageName = lastStageInfo.map(_.name).getOrElse("(Unknown Stage Name)")
      val lastStageDescription = lastStageData.flatMap(_.description).getOrElse("")
      val duration: Option[Long] = {
        job.startTime.map { start =>
          val end = job.endTime.getOrElse(System.currentTimeMillis())
          end - start
        }
      }
      val formattedDuration = duration.map(d => UIUtils.formatDuration(d)).getOrElse("Unknown")
      val formattedSubmissionTime = job.startTime.map(UIUtils.formatDate).getOrElse("Unknown")
      val detailUrl =
        "%s/jobs/job?id=%s".format(UIUtils.prependBaseUri(parent.basePath), job.jobId)
      <tr>
        <td sorttable_customkey={job.jobId.toString}>
          {job.jobId} {job.jobGroup.map(id => s"($id)").getOrElse("")}
        </td>
        <td>
          <div><em>{lastStageDescription}</em></div>
          <a href={detailUrl}>{lastStageName}</a>
        </td>
        <td sorttable_customkey={job.startTime.getOrElse(-1).toString}>
          {formattedSubmissionTime}
        </td>
        <td sorttable_customkey={duration.getOrElse(-1).toString}>{formattedDuration}</td>
        <td class="stage-progress-cell">
          {job.completedStageIndices.size}/{job.stageIds.size - job.numSkippedStages}
          {if (job.numFailedStages > 0) s"(${job.numFailedStages} failed)"}
          {if (job.numSkippedStages > 0) s"(${job.numSkippedStages} skipped)"}
        </td>
        <td class="progress-cell">
          {UIUtils.makeProgressBar(started = job.numActiveTasks, completed = job.numCompletedTasks,
           failed = job.numFailedTasks, skipped = job.numSkippedTasks,
           total = job.numTasks - job.numSkippedTasks)}
        </td>
      </tr>
    } 

private[spark] abstract class WebUITab(parent: WebUI, val prefix: String) {
  val pages = ArrayBuffer[WebUIPage]()
  val name = prefix.capitalize

  /** Attach a page to this tab. This prepends the page‘s prefix with the tab‘s own prefix. */
  def attachPage(page: WebUIPage) {
    page.prefix = (prefix + "/" + page.prefix).stripSuffix("/")
    pages += page
  }

  /** Get a list of header tabs from the parent UI. */
  def headerTabs: Seq[WebUITab] = parent.getTabs

  def basePath: String = parent.getBasePath
}

private[spark] abstract class WebUI( securityManager: SecurityManager, port: Int,
    conf: SparkConf, basePath: String = "", name: String = "") extends Logging {

  protected val tabs = ArrayBuffer[WebUITab]()
  protected val handlers = ArrayBuffer[ServletContextHandler]()
  protected var serverInfo: Option[ServerInfo] = None
  protected val localHostName = Utils.localHostName()
  protected val publicHostName = Option(System.getenv("SPARK_PUBLIC_DNS")).getOrElse(localHostName)
  private val className = Utils.getFormattedClassName(this)

  def getBasePath: String = basePath
  def getTabs: Seq[WebUITab] = tabs.toSeq
  def getHandlers: Seq[ServletContextHandler] = handlers.toSeq
  def getSecurityManager: SecurityManager = securityManager

  /** Attach a tab to this UI, along with all of its attached pages. */
  def attachTab(tab: WebUITab) {
    tab.pages.foreach(attachPage)
    tabs += tab
  }

  /** Attach a page to this UI. */
  def attachPage(page: WebUIPage) {
    val pagePath = "/" + page.prefix
    attachHandler(createServletHandler(pagePath,
      (request: HttpServletRequest) => page.render(request), securityManager, basePath))
    attachHandler(createServletHandler(pagePath.stripSuffix("/") + "/json",
      (request: HttpServletRequest) => page.renderJson(request), securityManager, basePath))
  }

  /** Attach a handler to this UI. */
  def attachHandler(handler: ServletContextHandler) {
    handlers += handler
    serverInfo.foreach { info =>
      info.rootHandler.addHandler(handler)
      if (!handler.isStarted) {
        handler.start()
      }
    }
  }

      serverInfo = Some(startJettyServer("0.0.0.0", port, handlers, conf, name))

  val hadoopConfiguration = SparkHadoopUtil.get.newConfiguration(conf)

private[spark] val executorMemory = conf.getOption("spark.executor.memory")
    .orElse(Option(System.getenv("SPARK_EXECUTOR_MEMORY")))
    .orElse(Option(System.getenv("SPARK_MEM")).map(warnSparkMem))
    .map(Utils.memoryStringToMb)
    .getOrElse(512)

  // Environment variables to pass to our executors.
  private[spark] val executorEnvs = HashMap[String, String]()

  for { (envKey, propKey) <- Seq(("SPARK_TESTING", "spark.testing"))
    value <- Option(System.getenv(envKey)).orElse(Option(System.getProperty(propKey)))} {
    executorEnvs(envKey) = value
  }
  Option(System.getenv("SPARK_PREPEND_CLASSES")).foreach { v =>
    executorEnvs("SPARK_PREPEND_CLASSES") = v
  }
  // The Mesos scheduler backend relies on this environment variable to set executor memory.
  executorEnvs("SPARK_EXECUTOR_MEMORY") = executorMemory + "m"
  executorEnvs ++= conf.getExecutorEnv

  // Set SPARK_USER for user who is running SparkContext.
  val sparkUser = Option {
    Option(System.getenv("SPARK_USER")).getOrElse(System.getProperty("user.name"))
  }.getOrElse {
    SparkContext.SPARK_UNKNOWN_USER
  }
  executorEnvs("SPARK_USER") = sparkUser 

  private[spark] var (schedulerBackend, taskScheduler) =
    SparkContext.createTaskScheduler(this, master)

    master match {
      case "local" =>
        val scheduler = new TaskSchedulerImpl(sc, MAX_LOCAL_TASK_FAILURES, isLocal = true)
        val backend = new LocalBackend(scheduler, 1)
        scheduler.initialize(backend)
        (backend, scheduler)

  var dagScheduler: DAGScheduler = null
  var backend: SchedulerBackend = null
  val mapOutputTracker = SparkEnv.get.mapOutputTracker
  var schedulableBuilder: SchedulableBuilder = null
  var rootPool: Pool = null
  // default scheduler is FIFO
  private val schedulingModeConf = conf.get("spark.scheduler.mode", "FIFO")
  val schedulingMode: SchedulingMode = try {
    SchedulingMode.withName(schedulingModeConf.toUpperCase)
  } catch {
    case e: java.util.NoSuchElementException =>
      throw new SparkException(s"Unrecognized spark.scheduler.mode: $schedulingModeConf")
  }

  // This is a var so that we can reset it for testing purposes.
  private[spark] var taskResultGetter = new TaskResultGetter(sc.env, this)

private[spark] class LocalBackend(scheduler: TaskSchedulerImpl, val totalCores: Int)
  extends SchedulerBackend with ExecutorBackend {

  private val appId = "local-" + System.currentTimeMillis
  var localActor: ActorRef = null

  override def start() {
    localActor = SparkEnv.get.actorSystem.actorOf(
      Props(new LocalActor(scheduler, this, totalCores)),
      "LocalBackendActor")
  }

  override def stop() {
    localActor ! StopExecutor
  }

  override def reviveOffers() {
    localActor ! ReviveOffers
  }

  override def defaultParallelism() =
    scheduler.conf.getInt("spark.default.parallelism", totalCores)

  override def killTask(taskId: Long, executorId: String, interruptThread: Boolean) {
    localActor ! KillTask(taskId, interruptThread)
  }

  override def statusUpdate(taskId: Long, state: TaskState, serializedData: ByteBuffer) {
    localActor ! StatusUpdate(taskId, state, serializedData)
  }

  override def applicationId(): String = appId
} 

def initialize(backend: SchedulerBackend) {
    this.backend = backend
    rootPool = new Pool("", schedulingMode, 0, 0)
    schedulableBuilder = {
      schedulingMode match {
        case SchedulingMode.FIFO =>
          new FIFOSchedulableBuilder(rootPool)
        case SchedulingMode.FAIR =>
          new FairSchedulableBuilder(rootPool, conf)
      }
    }
    schedulableBuilder.buildPools()
  } 

@volatile private[spark] var dagScheduler: DAGScheduler = _
    dagScheduler = new DAGScheduler(this)

  private[scheduler] val nextJobId = new AtomicInteger(0)
  private[scheduler] def numTotalJobs: Int = nextJobId.get()
  private val nextStageId = new AtomicInteger(0)

  private[scheduler] val jobIdToStageIds = new HashMap[Int, HashSet[Int]]
  private[scheduler] val stageIdToStage = new HashMap[Int, Stage]
  private[scheduler] val shuffleToMapStage = new HashMap[Int, Stage]
  private[scheduler] val jobIdToActiveJob = new HashMap[Int, ActiveJob]

  // Stages we need to run whose parents aren‘t done
  private[scheduler] val waitingStages = new HashSet[Stage]
  // Stages we are running right now
  private[scheduler] val runningStages = new HashSet[Stage]
  // Stages that must be resubmitted due to fetch failures
  private[scheduler] val failedStages = new HashSet[Stage]

  private[scheduler] val activeJobs = new HashSet[ActiveJob]

 // Contains the locations that each RDD‘s partitions are cached on
  private val cacheLocs = new HashMap[Int, Array[Seq[TaskLocation]]]
  private val failedEpoch = new HashMap[String, Long]

  private val dagSchedulerActorSupervisor =
    env.actorSystem.actorOf(Props(new DAGSchedulerActorSupervisor(this)))

  private val closureSerializer = SparkEnv.get.closureSerializer.newInstance()

  private[scheduler] var eventProcessActor: ActorRef = _
  private def initializeEventProcessActor() {
    // blocking the thread until supervisor is started, which ensures eventProcessActor is
    // not null before any job is submitted
    implicit val timeout = Timeout(30 seconds)
    val initEventActorReply =
      dagSchedulerActorSupervisor ? Props(new DAGSchedulerEventProcessActor(this))
    eventProcessActor = Await.result(initEventActorReply, timeout.duration).
      asInstanceOf[ActorRef]
  }

  initializeEventProcessActor() 

private[scheduler] class DAGSchedulerActorSupervisor(dagScheduler: DAGScheduler)
  extends Actor with Logging {

  override val supervisorStrategy =
    OneForOneStrategy() {
      case x: Exception =>
        logError("eventProcesserActor failed; shutting down SparkContext", x)
        try {
          dagScheduler.doCancelAllJobs()
        } catch {
          case t: Throwable => logError("DAGScheduler failed to cancel all jobs.", t)
        }
        dagScheduler.sc.stop()
        Stop
    }

  def receive = {
    case p: Props => sender ! context.actorOf(p)
    case _ => logWarning("received unknown message in DAGSchedulerActorSupervisor")
  }
} 

private[scheduler] class DAGSchedulerEventProcessActor(dagScheduler: DAGScheduler)
  extends Actor with Logging {
  override def preStart() {
    dagScheduler.taskScheduler.setDAGScheduler(dagScheduler)
  }
  /**
   * The main event loop of the DAG scheduler.
   */
  def receive = {
    case JobSubmitted(jobId, rdd, func, partitions, allowLocal, callSite, listener, properties) =>
      dagScheduler.handleJobSubmitted(jobId, rdd, func, partitions, allowLocal, callSite,
        listener, properties)
    case StageCancelled(stageId) =>
      dagScheduler.handleStageCancellation(stageId)
    case JobCancelled(jobId) =>
      dagScheduler.handleJobCancellation(jobId)
    case JobGroupCancelled(groupId) =>
      dagScheduler.handleJobGroupCancelled(groupId)
    case AllJobsCancelled =>
      dagScheduler.doCancelAllJobs()
    case ExecutorAdded(execId, host) =>
      dagScheduler.handleExecutorAdded(execId, host)
    case ExecutorLost(execId) =>
      dagScheduler.handleExecutorLost(execId, fetchFailed = false)
    case BeginEvent(task, taskInfo) =>
      dagScheduler.handleBeginEvent(task, taskInfo)
    case GettingResultEvent(taskInfo) =>
      dagScheduler.handleGetTaskResult(taskInfo)
    case completion @ CompletionEvent(task, reason, _, _, taskInfo, taskMetrics) =>
      dagScheduler.handleTaskCompletion(completion)
    case TaskSetFailed(taskSet, reason) =>
      dagScheduler.handleTaskSetFailed(taskSet, reason)
    case ResubmitFailedStages =>
      dagScheduler.resubmitFailedStages()
  }
  override def postStop() {
    // Cancel any active jobs in postStop hook
    dagScheduler.cleanUpAfterSchedulerStop()
  }