spark HealthTracker 源码

spark HealthTracker 代码

文件路径：/core/src/main/scala/org/apache/spark/scheduler/HealthTracker.scala

/*
 * Licensed to the Apache Software Foundation (ASF) under one or more
 * contributor license agreements.  See the NOTICE file distributed with
 * this work for additional information regarding copyright ownership.
 * The ASF licenses this file to You under the Apache License, Version 2.0
 * (the "License"); you may not use this file except in compliance with
 * the License.  You may obtain a copy of the License at
 *
 *    http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

package org.apache.spark.scheduler

import java.util.concurrent.atomic.AtomicReference

import scala.collection.mutable.{ArrayBuffer, HashMap, HashSet}

import org.apache.spark.{ExecutorAllocationClient, SparkConf, SparkContext}
import org.apache.spark.internal.Logging
import org.apache.spark.internal.config
import org.apache.spark.util.{Clock, SystemClock, Utils}

/**
 * HealthTracker is designed to track problematic executors and nodes.  It supports excluding
 * executors and nodes across an entire application (with a periodic expiry). TaskSetManagers add
 * additional logic for exclusion of executors and nodes for individual tasks and stages which
 * works in concert with the logic here.
 *
 * The tracker needs to deal with a variety of workloads, e.g.:
 *
 *  * bad user code -- this may lead to many task failures, but that should not count against
 *      individual executors
 *  * many small stages -- this may prevent a bad executor for having many failures within one
 *      stage, but still many failures over the entire application
 *  * "flaky" executors -- they don't fail every task, but are still faulty enough to merit
 *      excluding
 *  * missing shuffle files -- may trigger fetch failures on healthy executors.
 *
 * See the design doc on SPARK-8425 for a more in-depth discussion. Note SPARK-32037 renamed
 * the feature.
 *
 * THREADING: As with most helpers of TaskSchedulerImpl, this is not thread-safe.  Though it is
 * called by multiple threads, callers must already have a lock on the TaskSchedulerImpl.  The
 * one exception is [[excludedNodeList()]], which can be called without holding a lock.
 */
private[scheduler] class HealthTracker (
    private val listenerBus: LiveListenerBus,
    conf: SparkConf,
    allocationClient: Option[ExecutorAllocationClient],
    clock: Clock = new SystemClock()) extends Logging {

  def this(sc: SparkContext, allocationClient: Option[ExecutorAllocationClient]) = {
    this(sc.listenerBus, sc.conf, allocationClient)
  }

  HealthTracker.validateExcludeOnFailureConfs(conf)
  private val MAX_FAILURES_PER_EXEC = conf.get(config.MAX_FAILURES_PER_EXEC)
  private val MAX_FAILED_EXEC_PER_NODE = conf.get(config.MAX_FAILED_EXEC_PER_NODE)
  val EXCLUDE_ON_FAILURE_TIMEOUT_MILLIS = HealthTracker.getExludeOnFailureTimeout(conf)
  private val EXCLUDE_FETCH_FAILURE_ENABLED =
    conf.get(config.EXCLUDE_ON_FAILURE_FETCH_FAILURE_ENABLED)
  private val EXCLUDE_ON_FAILURE_DECOMMISSION_ENABLED =
    conf.get(config.EXCLUDE_ON_FAILURE_DECOMMISSION_ENABLED)

  /**
   * A map from executorId to information on task failures. Tracks the time of each task failure,
   * so that we can avoid excluding executors due to failures that are very far apart. We do not
   * actively remove from this as soon as tasks hit their timeouts, to avoid the time it would take
   * to do so. But it will not grow too large, because as soon as an executor gets too many
   * failures, we exclude the executor and remove its entry here.
   */
  private val executorIdToFailureList = new HashMap[String, ExecutorFailureList]()
  val executorIdToExcludedStatus = new HashMap[String, ExcludedExecutor]()
  val nodeIdToExcludedExpiryTime = new HashMap[String, Long]()
  /**
   * An immutable copy of the set of nodes that are currently excluded.  Kept in an
   * AtomicReference to make [[excludedNodeList()]] thread-safe.
   */
  private val _excludedNodeList = new AtomicReference[Set[String]](Set())
  /**
   * Time when the next excluded node will expire.  Used as a shortcut to
   * avoid iterating over all entries in the excludedNodeList when none will have expired.
   */
  var nextExpiryTime: Long = Long.MaxValue
  /**
   * Mapping from nodes to all of the executors that have been excluded on that node. We do *not*
   * remove from this when executors are removed from spark, so we can track when we get multiple
   * successive excluded executors on one node.  Nonetheless, it will not grow too large because
   * there cannot be many excluded executors on one node, before we stop requesting more
   * executors on that node, and we clean up the list of exluded executors once an executor has
   * been excluded for EXCLUDE_ON_FAILURE_TIMEOUT_MILLIS.
   */
  val nodeToExcludedExecs = new HashMap[String, HashSet[String]]()

  /**
   * Include executors and nodes that have been excluded for at least
   * EXCLUDE_ON_FAILURE_TIMEOUT_MILLIS
   */
  def applyExcludeOnFailureTimeout(): Unit = {
    val now = clock.getTimeMillis()
    // quickly check if we've got anything to expire that is excluded -- if not,
    // avoid doing any work
    if (now > nextExpiryTime) {
      // Apply the timeout to excluded nodes and executors
      val execsToInclude = executorIdToExcludedStatus.filter(_._2.expiryTime < now).keys
      if (execsToInclude.nonEmpty) {
        // Include any executors that have been exluded longer than the excludeOnFailure timeout.
        logInfo(s"Removing executors $execsToInclude from exclude list because the " +
          s"the executors have reached the timed out")
        execsToInclude.foreach { exec =>
          val status = executorIdToExcludedStatus.remove(exec).get
          val failedExecsOnNode = nodeToExcludedExecs(status.node)
          // post both to keep backwards compatibility
          listenerBus.post(SparkListenerExecutorUnblacklisted(now, exec))
          listenerBus.post(SparkListenerExecutorUnexcluded(now, exec))
          failedExecsOnNode.remove(exec)
          if (failedExecsOnNode.isEmpty) {
            nodeToExcludedExecs.remove(status.node)
          }
        }
      }
      val nodesToInclude = nodeIdToExcludedExpiryTime.filter(_._2 < now).keys
      if (nodesToInclude.nonEmpty) {
        // Include any nodes that have been excluded longer than the excludeOnFailure timeout.
        logInfo(s"Removing nodes $nodesToInclude from exclude list because the " +
          s"nodes have reached has timed out")
        nodesToInclude.foreach { node =>
          nodeIdToExcludedExpiryTime.remove(node)
          // post both to keep backwards compatibility
          listenerBus.post(SparkListenerNodeUnblacklisted(now, node))
          listenerBus.post(SparkListenerNodeUnexcluded(now, node))
        }
        _excludedNodeList.set(nodeIdToExcludedExpiryTime.keySet.toSet)
      }
      updateNextExpiryTime()
    }
  }

  private def updateNextExpiryTime(): Unit = {
    val execMinExpiry = if (executorIdToExcludedStatus.nonEmpty) {
      executorIdToExcludedStatus.map{_._2.expiryTime}.min
    } else {
      Long.MaxValue
    }
    val nodeMinExpiry = if (nodeIdToExcludedExpiryTime.nonEmpty) {
      nodeIdToExcludedExpiryTime.values.min
    } else {
      Long.MaxValue
    }
    nextExpiryTime = math.min(execMinExpiry, nodeMinExpiry)
  }

  private def killExecutor(exec: String, msg: String): Unit = {
    val fullMsg = if (EXCLUDE_ON_FAILURE_DECOMMISSION_ENABLED) {
      s"${msg} (actually decommissioning)"
    } else {
      msg
    }
    allocationClient match {
      case Some(a) =>
        logInfo(fullMsg)
        if (EXCLUDE_ON_FAILURE_DECOMMISSION_ENABLED) {
          a.decommissionExecutor(exec, ExecutorDecommissionInfo(fullMsg),
            adjustTargetNumExecutors = false)
        } else {
          a.killExecutors(Seq(exec), adjustTargetNumExecutors = false, countFailures = false,
            force = true)
        }
      case None =>
        logInfo(s"Not attempting to kill excluded executor id $exec " +
          s"since allocation client is not defined.")
    }
  }

  private def killExcludedExecutor(exec: String): Unit = {
    if (conf.get(config.EXCLUDE_ON_FAILURE_KILL_ENABLED)) {
      killExecutor(exec, s"Killing excluded executor id $exec since " +
        s"${config.EXCLUDE_ON_FAILURE_KILL_ENABLED.key} is set.")
    }
  }

  private[scheduler] def killExcludedIdleExecutor(exec: String): Unit = {
    killExecutor(exec,
      s"Killing excluded idle executor id $exec because of task unschedulability and trying " +
        "to acquire a new executor.")
  }

  private def killExecutorsOnExcludedNode(node: String): Unit = {
    if (conf.get(config.EXCLUDE_ON_FAILURE_KILL_ENABLED)) {
      allocationClient match {
        case Some(a) =>
          if (EXCLUDE_ON_FAILURE_DECOMMISSION_ENABLED) {
            logInfo(s"Decommissioning all executors on excluded host $node " +
              s"since ${config.EXCLUDE_ON_FAILURE_KILL_ENABLED.key} is set.")
            if (!a.decommissionExecutorsOnHost(node)) {
              logError(s"Decommissioning executors on $node failed.")
            }
          } else {
            logInfo(s"Killing all executors on excluded host $node " +
              s"since ${config.EXCLUDE_ON_FAILURE_KILL_ENABLED.key} is set.")
            if (!a.killExecutorsOnHost(node)) {
              logError(s"Killing executors on node $node failed.")
            }
          }
        case None =>
          logWarning(s"Not attempting to kill executors on excluded host $node " +
            s"since allocation client is not defined.")
      }
    }
  }

  def updateExcludedForFetchFailure(host: String, exec: String): Unit = {
    if (EXCLUDE_FETCH_FAILURE_ENABLED) {
      // If we exclude on fetch failures, we are implicitly saying that we believe the failure is
      // non-transient, and can't be recovered from (even if this is the first fetch failure,
      // stage is retried after just one failure, so we don't always get a chance to collect
      // multiple fetch failures).
      // If the external shuffle-service is on, then every other executor on this node would
      // be suffering from the same issue, so we should exclude (and potentially kill) all
      // of them immediately.

      val now = clock.getTimeMillis()
      val expiryTimeForNewExcludes = now + EXCLUDE_ON_FAILURE_TIMEOUT_MILLIS

      if (conf.get(config.SHUFFLE_SERVICE_ENABLED)) {
        if (!nodeIdToExcludedExpiryTime.contains(host)) {
          logInfo(s"excluding node $host due to fetch failure of external shuffle service")

          nodeIdToExcludedExpiryTime.put(host, expiryTimeForNewExcludes)
          // post both to keep backwards compatibility
          listenerBus.post(SparkListenerNodeBlacklisted(now, host, 1))
          listenerBus.post(SparkListenerNodeExcluded(now, host, 1))
          _excludedNodeList.set(nodeIdToExcludedExpiryTime.keySet.toSet)
          killExecutorsOnExcludedNode(host)
          updateNextExpiryTime()
        }
      } else if (!executorIdToExcludedStatus.contains(exec)) {
        logInfo(s"Excluding executor $exec due to fetch failure")

        executorIdToExcludedStatus.put(exec, ExcludedExecutor(host, expiryTimeForNewExcludes))
        // We hardcoded number of failure tasks to 1 for fetch failure, because there's no
        // reattempt for such failure.
        // post both to keep backwards compatibility
        listenerBus.post(SparkListenerExecutorBlacklisted(now, exec, 1))
        listenerBus.post(SparkListenerExecutorExcluded(now, exec, 1))
        updateNextExpiryTime()
        killExcludedExecutor(exec)

        val excludedExecsOnNode = nodeToExcludedExecs.getOrElseUpdate(host, HashSet[String]())
        excludedExecsOnNode += exec
      }
    }
  }

  def updateExcludedForSuccessfulTaskSet(
      stageId: Int,
      stageAttemptId: Int,
      failuresByExec: HashMap[String, ExecutorFailuresInTaskSet]): Unit = {
    // if any tasks failed, we count them towards the overall failure count for the executor at
    // this point.
    val now = clock.getTimeMillis()
    failuresByExec.foreach { case (exec, failuresInTaskSet) =>
      val appFailuresOnExecutor =
        executorIdToFailureList.getOrElseUpdate(exec, new ExecutorFailureList)
      appFailuresOnExecutor.addFailures(stageId, stageAttemptId, failuresInTaskSet)
      appFailuresOnExecutor.dropFailuresWithTimeoutBefore(now)
      val newTotal = appFailuresOnExecutor.numUniqueTaskFailures

      val expiryTimeForNewExcludes = now + EXCLUDE_ON_FAILURE_TIMEOUT_MILLIS
      // If this pushes the total number of failures over the threshold, exclude the executor.
      // If its already excluded, we avoid "re-excluding" (which can happen if there were
      // other tasks already running in another taskset when it got excluded), because it makes
      // some of the logic around expiry times a little more confusing.  But it also wouldn't be a
      // problem to re-exclude, with a later expiry time.
      if (newTotal >= MAX_FAILURES_PER_EXEC && !executorIdToExcludedStatus.contains(exec)) {
        logInfo(s"Excluding executor id: $exec because it has $newTotal" +
          s" task failures in successful task sets")
        val node = failuresInTaskSet.node
        executorIdToExcludedStatus.put(exec, ExcludedExecutor(node, expiryTimeForNewExcludes))
        // post both to keep backwards compatibility
        listenerBus.post(SparkListenerExecutorBlacklisted(now, exec, newTotal))
        listenerBus.post(SparkListenerExecutorExcluded(now, exec, newTotal))
        executorIdToFailureList.remove(exec)
        updateNextExpiryTime()
        killExcludedExecutor(exec)

        // In addition to excluding the executor, we also update the data for failures on the
        // node, and potentially exclude the entire node as well.
        val excludedExecsOnNode = nodeToExcludedExecs.getOrElseUpdate(node, HashSet[String]())
        excludedExecsOnNode += exec
        // If the node is already excluded, we avoid adding it again with a later expiry
        // time.
        if (excludedExecsOnNode.size >= MAX_FAILED_EXEC_PER_NODE &&
            !nodeIdToExcludedExpiryTime.contains(node)) {
          logInfo(s"Excluding node $node because it has ${excludedExecsOnNode.size} " +
            s"executors excluded: ${excludedExecsOnNode}")
          nodeIdToExcludedExpiryTime.put(node, expiryTimeForNewExcludes)
          // post both to keep backwards compatibility
          listenerBus.post(SparkListenerNodeBlacklisted(now, node, excludedExecsOnNode.size))
          listenerBus.post(SparkListenerNodeExcluded(now, node, excludedExecsOnNode.size))
          _excludedNodeList.set(nodeIdToExcludedExpiryTime.keySet.toSet)
          killExecutorsOnExcludedNode(node)
        }
      }
    }
  }

  def isExecutorExcluded(executorId: String): Boolean = {
    executorIdToExcludedStatus.contains(executorId)
  }

  /**
   * Get the full set of nodes that are excluded.  Unlike other methods in this class, this *IS*
   * thread-safe -- no lock required on a taskScheduler.
   */
  def excludedNodeList(): Set[String] = {
    _excludedNodeList.get()
  }

  def isNodeExcluded(node: String): Boolean = {
    nodeIdToExcludedExpiryTime.contains(node)
  }

  def handleRemovedExecutor(executorId: String): Unit = {
    // We intentionally do not clean up executors that are already excluded in
    // nodeToExcludedExecs, so that if another executor on the same node gets excluded, we can
    // exclude the entire node. We also can't clean up executorIdToExcludedStatus, so we can
    // eventually remove the executor after the timeout. Despite not clearing those structures
    // here, we don't expect they will grow too big since you won't get too many executors on one
    // node, and the timeout will clear it up periodically in any case.
    executorIdToFailureList -= executorId
  }

  /**
   * Tracks all failures for one executor (that have not passed the timeout).
   *
   * In general we actually expect this to be extremely small, since it won't contain more than the
   * maximum number of task failures before an executor is failed (default 2).
   */
  private[scheduler] final class ExecutorFailureList extends Logging {

    private case class TaskId(stage: Int, stageAttempt: Int, taskIndex: Int)

    /**
     * All failures on this executor in successful task sets.
     */
    private var failuresAndExpiryTimes = ArrayBuffer[(TaskId, Long)]()
    /**
     * As an optimization, we track the min expiry time over all entries in failuresAndExpiryTimes
     * so its quick to tell if there are any failures with expiry before the current time.
     */
    private var minExpiryTime = Long.MaxValue

    def addFailures(
        stage: Int,
        stageAttempt: Int,
        failuresInTaskSet: ExecutorFailuresInTaskSet): Unit = {
      failuresInTaskSet.taskToFailureCountAndFailureTime.foreach {
        case (taskIdx, (_, failureTime)) =>
          val expiryTime = failureTime + EXCLUDE_ON_FAILURE_TIMEOUT_MILLIS
          failuresAndExpiryTimes += ((TaskId(stage, stageAttempt, taskIdx), expiryTime))
          if (expiryTime < minExpiryTime) {
            minExpiryTime = expiryTime
          }
      }
    }

    /**
     * The number of unique tasks that failed on this executor.  Only counts failures within the
     * timeout, and in successful tasksets.
     */
    def numUniqueTaskFailures: Int = failuresAndExpiryTimes.size

    def isEmpty: Boolean = failuresAndExpiryTimes.isEmpty

    /**
     * Apply the timeout to individual tasks.  This is to prevent one-off failures that are very
     * spread out in time (and likely have nothing to do with problems on the executor) from
     * triggering exlusion.  However, note that we do *not* remove executors and nodes from
     * being excluded as we expire individual task failures -- each have their own timeout.  E.g.,
     * suppose:
     *  * timeout = 10, maxFailuresPerExec = 2
     *  * Task 1 fails on exec 1 at time 0
     *  * Task 2 fails on exec 1 at time 5
     * -->  exec 1 is excluded from time 5 - 15.
     * This is to simplify the implementation, as well as keep the behavior easier to understand
     * for the end user.
     */
    def dropFailuresWithTimeoutBefore(dropBefore: Long): Unit = {
      if (minExpiryTime < dropBefore) {
        var newMinExpiry = Long.MaxValue
        val newFailures = new ArrayBuffer[(TaskId, Long)]
        failuresAndExpiryTimes.foreach { case (task, expiryTime) =>
          if (expiryTime >= dropBefore) {
            newFailures += ((task, expiryTime))
            if (expiryTime < newMinExpiry) {
              newMinExpiry = expiryTime
            }
          }
        }
        failuresAndExpiryTimes = newFailures
        minExpiryTime = newMinExpiry
      }
    }

    override def toString(): String = {
      s"failures = $failuresAndExpiryTimes"
    }
  }

}

private[spark] object HealthTracker extends Logging {

  private val DEFAULT_TIMEOUT = "1h"

  /**
   * Returns true if the excludeOnFailure is enabled, based on checking the configuration
   * in the following order:
   * 1. Is it specifically enabled or disabled?
   * 2. Is it enabled via the legacy timeout conf?
   * 3. Default is off
   */
  def isExcludeOnFailureEnabled(conf: SparkConf): Boolean = {
    conf.get(config.EXCLUDE_ON_FAILURE_ENABLED) match {
      case Some(enabled) =>
        enabled
      case None =>
        // if they've got a non-zero setting for the legacy conf, always enable it,
        // otherwise, use the default.
        val legacyKey = config.EXCLUDE_ON_FAILURE_LEGACY_TIMEOUT_CONF.key
        conf.get(config.EXCLUDE_ON_FAILURE_LEGACY_TIMEOUT_CONF).exists { legacyTimeout =>
          if (legacyTimeout == 0) {
            logWarning(s"Turning off excludeOnFailure due to legacy configuration: $legacyKey == 0")
            false
          } else {
            logWarning(s"Turning on excludeOnFailure due to legacy configuration: $legacyKey > 0")
            true
          }
        }
    }
  }

  def getExludeOnFailureTimeout(conf: SparkConf): Long = {
    conf.get(config.EXCLUDE_ON_FAILURE_TIMEOUT_CONF).getOrElse {
      conf.get(config.EXCLUDE_ON_FAILURE_LEGACY_TIMEOUT_CONF).getOrElse {
        Utils.timeStringAsMs(DEFAULT_TIMEOUT)
      }
    }
  }

  /**
   * Verify that exclude on failure configurations are consistent; if not, throw an exception.
   * Should only be called if excludeOnFailure is enabled.
   *
   * The configuration is expected to adhere to a few invariants.  Default values
   * follow these rules of course, but users may unwittingly change one configuration
   * without making the corresponding adjustment elsewhere. This ensures we fail-fast when
   * there are such misconfigurations.
   */
  def validateExcludeOnFailureConfs(conf: SparkConf): Unit = {

    def mustBePos(k: String, v: String): Unit = {
      throw new IllegalArgumentException(s"$k was $v, but must be > 0.")
    }

    Seq(
      config.MAX_TASK_ATTEMPTS_PER_EXECUTOR,
      config.MAX_TASK_ATTEMPTS_PER_NODE,
      config.MAX_FAILURES_PER_EXEC_STAGE,
      config.MAX_FAILED_EXEC_PER_NODE_STAGE,
      config.MAX_FAILURES_PER_EXEC,
      config.MAX_FAILED_EXEC_PER_NODE
    ).foreach { config =>
      val v = conf.get(config)
      if (v <= 0) {
        mustBePos(config.key, v.toString)
      }
    }

    val timeout = getExludeOnFailureTimeout(conf)
    if (timeout <= 0) {
      // first, figure out where the timeout came from, to include the right conf in the message.
      conf.get(config.EXCLUDE_ON_FAILURE_TIMEOUT_CONF) match {
        case Some(t) =>
          mustBePos(config.EXCLUDE_ON_FAILURE_TIMEOUT_CONF.key, timeout.toString)
        case None =>
          mustBePos(config.EXCLUDE_ON_FAILURE_LEGACY_TIMEOUT_CONF.key, timeout.toString)
      }
    }

    val maxTaskFailures = conf.get(config.TASK_MAX_FAILURES)
    val maxNodeAttempts = conf.get(config.MAX_TASK_ATTEMPTS_PER_NODE)

    if (maxNodeAttempts >= maxTaskFailures) {
      throw new IllegalArgumentException(s"${config.MAX_TASK_ATTEMPTS_PER_NODE.key} " +
        s"( = ${maxNodeAttempts}) was >= ${config.TASK_MAX_FAILURES.key} " +
        s"( = ${maxTaskFailures} ). Though excludeOnFailure is enabled, with this configuration, " +
        s"Spark will not be robust to one bad node. Decrease " +
        s"${config.MAX_TASK_ATTEMPTS_PER_NODE.key}, increase ${config.TASK_MAX_FAILURES.key}, " +
        s"or disable excludeOnFailure with ${config.EXCLUDE_ON_FAILURE_ENABLED.key}")
    }
  }
}

private final case class ExcludedExecutor(node: String, expiryTime: Long)

相关信息

spark 源码目录

相关文章

spark AccumulableInfo 源码

spark ActiveJob 源码

spark AsyncEventQueue 源码

spark BarrierJobAllocationFailed 源码

spark DAGScheduler 源码

spark DAGSchedulerEvent 源码

spark DAGSchedulerSource 源码

spark EventLoggingListener 源码

spark ExecutorDecommissionInfo 源码

spark ExecutorFailuresInTaskSet 源码