spark YarnAllocatorNodeHealthTracker 源码

spark YarnAllocatorNodeHealthTracker 代码

文件路径：/resource-managers/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocatorNodeHealthTracker.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.deploy.yarn

import scala.collection.JavaConverters._
import scala.collection.mutable
import scala.collection.mutable.HashMap

import org.apache.hadoop.yarn.client.api.AMRMClient
import org.apache.hadoop.yarn.client.api.AMRMClient.ContainerRequest

import org.apache.spark.SparkConf
import org.apache.spark.deploy.yarn.config._
import org.apache.spark.internal.Logging
import org.apache.spark.internal.config._
import org.apache.spark.scheduler.HealthTracker
import org.apache.spark.util.{Clock, SystemClock}

/**
 * YarnAllocatorNodeHealthTracker is responsible for tracking the health of nodes
 * and synchronizing the node list to YARN as to which nodes are excluded.
 *
 * Excluding nodes are coming from two different sources:
 *
 * <ul>
 *   <li> from the scheduler as task level excluded nodes
 *   <li> from this class (tracked here) as YARN resource allocation problems
 * </ul>
 *
 * The reason to realize this logic here (and not in the driver) is to avoid possible delays
 * between synchronizing the excluded nodes with YARN and resource allocations.
 */
private[spark] class YarnAllocatorNodeHealthTracker(
    sparkConf: SparkConf,
    amClient: AMRMClient[ContainerRequest],
    failureTracker: FailureTracker)
  extends Logging {

  private val excludeOnFailureTimeoutMillis = HealthTracker.getExludeOnFailureTimeout(sparkConf)

  private val launchExcludeOnFailureEnabled =
    sparkConf.get(YARN_EXECUTOR_LAUNCH_EXCLUDE_ON_FAILURE_ENABLED)

  private val maxFailuresPerHost = sparkConf.get(MAX_FAILED_EXEC_PER_NODE)

  private val excludeNodes = sparkConf.get(YARN_EXCLUDE_NODES).toSet

  private val allocatorExcludedNodeList = new HashMap[String, Long]()

  private var currentExcludededYarnNodes = Set.empty[String]

  private var schedulerExcludedNodeList = Set.empty[String]

  private var numClusterNodes = Int.MaxValue

  def setNumClusterNodes(numClusterNodes: Int): Unit = {
    this.numClusterNodes = numClusterNodes
  }

  def handleResourceAllocationFailure(hostOpt: Option[String]): Unit = {
    hostOpt match {
      case Some(hostname) if launchExcludeOnFailureEnabled =>
        // failures on an already excluded node are not even tracked.
        // otherwise, such failures could shutdown the application
        // as resource requests are asynchronous
        // and a late failure response could exceed MAX_EXECUTOR_FAILURES
        if (!schedulerExcludedNodeList.contains(hostname) &&
            !allocatorExcludedNodeList.contains(hostname)) {
          failureTracker.registerFailureOnHost(hostname)
          updateAllocationExcludedNodes(hostname)
        }
      case _ =>
        failureTracker.registerExecutorFailure()
    }
  }

  private def updateAllocationExcludedNodes(hostname: String): Unit = {
    val failuresOnHost = failureTracker.numFailuresOnHost(hostname)
    if (failuresOnHost > maxFailuresPerHost) {
      logInfo(s"excluding $hostname as YARN allocation failed $failuresOnHost times")
      allocatorExcludedNodeList.put(
        hostname,
        failureTracker.clock.getTimeMillis() + excludeOnFailureTimeoutMillis)
      refreshExcludedNodes()
    }
  }

  def setSchedulerExcludedNodes(schedulerExcludedNodesWithExpiry: Set[String]): Unit = {
    this.schedulerExcludedNodeList = schedulerExcludedNodesWithExpiry
    refreshExcludedNodes()
  }

  def isAllNodeExcluded: Boolean = {
    if (numClusterNodes <= 0) {
      logWarning("No available nodes reported, please check Resource Manager.")
      false
    } else {
      currentExcludededYarnNodes.size >= numClusterNodes
    }
  }

  private def refreshExcludedNodes(): Unit = {
    removeExpiredYarnExcludedNodes()
    val allExcludedNodes =
      excludeNodes ++ schedulerExcludedNodeList ++ allocatorExcludedNodeList.keySet
    synchronizeExcludedNodesWithYarn(allExcludedNodes)
  }

  private def synchronizeExcludedNodesWithYarn(nodesToExclude: Set[String]): Unit = {
    // Update YARN with the nodes that are excluded for this application,
    // in order to avoid allocating new Containers on the problematic nodes.
    val additions = (nodesToExclude -- currentExcludededYarnNodes).toList.sorted
    val removals = (currentExcludededYarnNodes -- nodesToExclude).toList.sorted
    if (additions.nonEmpty) {
      logInfo(s"adding nodes to YARN application master's excluded node list: $additions")
    }
    if (removals.nonEmpty) {
      logInfo(s"removing nodes from YARN application master's excluded node list: $removals")
    }
    if (additions.nonEmpty || removals.nonEmpty) {
      // Note YARNs api for excluding nodes is updateBlacklist.
      // TODO - We need to update once Hadoop changes -
      // https://issues.apache.org/jira/browse/HADOOP-17169
      amClient.updateBlacklist(additions.asJava, removals.asJava)
    }
    currentExcludededYarnNodes = nodesToExclude
  }

  private def removeExpiredYarnExcludedNodes(): Unit = {
    val now = failureTracker.clock.getTimeMillis()
    allocatorExcludedNodeList.retain { (_, expiryTime) => expiryTime > now }
  }
}

/**
 * FailureTracker is responsible for tracking executor failures both for each host separately
 * and for all hosts altogether.
 */
private[spark] class FailureTracker(
    sparkConf: SparkConf,
    val clock: Clock = new SystemClock) extends Logging {

  private val executorFailuresValidityInterval =
    sparkConf.get(config.EXECUTOR_ATTEMPT_FAILURE_VALIDITY_INTERVAL_MS).getOrElse(-1L)

  // Queue to store the timestamp of failed executors for each host
  private val failedExecutorsTimeStampsPerHost = mutable.Map[String, mutable.Queue[Long]]()

  private val failedExecutorsTimeStamps = new mutable.Queue[Long]()

  private def updateAndCountFailures(failedExecutorsWithTimeStamps: mutable.Queue[Long]): Int = {
    val endTime = clock.getTimeMillis()
    while (executorFailuresValidityInterval > 0 &&
        failedExecutorsWithTimeStamps.nonEmpty &&
        failedExecutorsWithTimeStamps.head < endTime - executorFailuresValidityInterval) {
      failedExecutorsWithTimeStamps.dequeue()
    }
    failedExecutorsWithTimeStamps.size
  }

  def numFailedExecutors: Int = synchronized {
    updateAndCountFailures(failedExecutorsTimeStamps)
  }

  def registerFailureOnHost(hostname: String): Unit = synchronized {
    val timeMillis = clock.getTimeMillis()
    failedExecutorsTimeStamps.enqueue(timeMillis)
    val failedExecutorsOnHost =
      failedExecutorsTimeStampsPerHost.getOrElse(hostname, {
        val failureOnHost = mutable.Queue[Long]()
        failedExecutorsTimeStampsPerHost.put(hostname, failureOnHost)
        failureOnHost
      })
    failedExecutorsOnHost.enqueue(timeMillis)
  }

  def registerExecutorFailure(): Unit = synchronized {
    val timeMillis = clock.getTimeMillis()
    failedExecutorsTimeStamps.enqueue(timeMillis)
  }

  def numFailuresOnHost(hostname: String): Int = {
    failedExecutorsTimeStampsPerHost.get(hostname).map { failedExecutorsOnHost =>
      updateAndCountFailures(failedExecutorsOnHost)
    }.getOrElse(0)
  }

}

相关信息

spark 源码目录

相关文章

spark ApplicationMaster 源码

spark ApplicationMasterArguments 源码

spark ApplicationMasterSource 源码

spark Client 源码

spark ClientArguments 源码

spark ClientDistributedCacheManager 源码

spark ExecutorRunnable 源码

spark LocalityPreferredContainerPlacementStrategy 源码

spark ResourceRequestHelper 源码

spark SparkRackResolver 源码