spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From andrewor14 <...@git.apache.org>
Subject [GitHub] spark pull request: [SPARK-4286] Integrate external shuffle servic...
Date Fri, 13 Feb 2015 04:45:53 GMT
Github user andrewor14 commented on a diff in the pull request:

    https://github.com/apache/spark/pull/3861#discussion_r24644199
  
    --- Diff: core/src/main/scala/org/apache/spark/executor/CoarseGrainedMesosExecutorBackend.scala
---
    @@ -0,0 +1,212 @@
    +/*
    + * 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.executor
    +
    +import org.apache.spark.{SparkConf, Logging, SecurityManager}
    +import org.apache.mesos.{Executor => MesosExecutor, ExecutorDriver, MesosExecutorDriver,
MesosNativeLibrary}
    +import org.apache.spark.util.{Utils, SignalLogger}
    +import org.apache.spark.deploy.SparkHadoopUtil
    +import org.apache.mesos.Protos._
    +import org.apache.spark.deploy.worker.StandaloneWorkerShuffleService
    +import scala.collection.JavaConversions._
    +import scala.io.Source
    +import java.io.{File, PrintWriter}
    +
    +/**
    + * The Coarse grained Mesos executor backend is responsible for launching the shuffle
service
    + * and the CoarseGrainedExecutorBackend actor.
    + * This is assuming the scheduler detected that the shuffle service is enabled and launches
    + * this class instead of CoarseGrainedExecutorBackend directly.
    + */
    +private[spark] class CoarseGrainedMesosExecutorBackend(val sparkConf: SparkConf)
    +  extends MesosExecutor
    +  with Logging {
    +
    +  private var shuffleService: StandaloneWorkerShuffleService = null
    +  private var driver: ExecutorDriver = null
    +  private var executorProc: Process = null
    +  private var taskId: TaskID = null
    +  @volatile var killed = false
    +
    +  override def registered(
    +      driver: ExecutorDriver,
    +      executorInfo: ExecutorInfo,
    +      frameworkInfo: FrameworkInfo,
    +      slaveInfo: SlaveInfo) {
    +    this.driver = driver
    +    logInfo("Coarse Grained Mesos Executor '" + executorInfo.getExecutorId.getValue +
    +      "' is registered.")
    +
    +    if (shuffleService == null) {
    +      sparkConf.set("spark.shuffle.service.enabled", "true")
    +      shuffleService = new StandaloneWorkerShuffleService(sparkConf, new SecurityManager(sparkConf))
    +      shuffleService.startIfEnabled()
    +    }
    +  }
    +
    +  override def launchTask(d: ExecutorDriver, taskInfo: TaskInfo) {
    +    if (executorProc != null) {
    +      logError("Received LaunchTask while executor is already running")
    +      val status = TaskStatus.newBuilder()
    +        .setTaskId(taskInfo.getTaskId)
    +        .setSlaveId(taskInfo.getSlaveId)
    +        .setState(TaskState.TASK_FAILED)
    +        .setMessage("Received LaunchTask while executor is already running")
    +        .build()
    +      d.sendStatusUpdate(status)
    +      return
    +    }
    +
    +    killed = false
    +
    +    logInfo("Launching task id: " + taskInfo.getTaskId.getValue)
    +
    +    // We are launching the CoarseGrainedExecutorBackend via subprocess
    +    // because the backend is designed to run in its own process.
    +    // Since it's a shared class we are preserving the existing behavior
    +    // and launching it as a subprocess here.
    +    val command = "exec " + Utils.deserialize[String](taskInfo.getData().toByteArray)
    +
    +    logInfo("Running command: " + command)
    +
    +    // Mesos only work on linux platforms, so we assume bash is available is Mesos is
used.
    +    val pb = new ProcessBuilder("/bin/bash", "-c", command)
    +
    +    val currentEnvVars = pb.environment()
    +    for (variable <- taskInfo.getExecutor.getCommand.getEnvironment.getVariablesList())
{
    +      currentEnvVars.put(variable.getName, variable.getValue)
    +    }
    +
    +    executorProc = pb.start()
    +
    +    new Thread("stderr reader for task " + taskInfo.getTaskId.getValue) {
    +      override def run() {
    +        for (line <- Source.fromInputStream(executorProc.getErrorStream).getLines)
{
    +          System.err.println(line)
    +        }
    +      }
    +    }.start()
    +
    +    new Thread("stdout reader for task " + taskInfo.getTaskId.getValue) {
    +      override def run() {
    +        for (line <- Source.fromInputStream(executorProc.getInputStream).getLines)
{
    +          System.out.println(line)
    +        }
    +      }
    +    }.start()
    +
    +    driver.sendStatusUpdate(TaskStatus.newBuilder()
    +      .setState(TaskState.TASK_RUNNING)
    +      .setTaskId(taskInfo.getTaskId)
    +      .build)
    +
    +    new Thread("process waiter for mesos executor for task " + taskInfo.getTaskId.getValue)
{
    +      override def run() {
    +        executorProc.waitFor()
    +        val (state, msg) = if (killed) {
    +          (TaskState.TASK_KILLED, "")
    +        }  else if (executorProc.exitValue() == 0) {
    +          (TaskState.TASK_FINISHED, "")
    +        } else {
    +          (TaskState.TASK_FAILED, "Exited with status: " + executorProc.exitValue().toString)
    +        }
    +        // We leave the shuffle service running after the task.
    +        cleanup(state, msg)
    +      }
    +    }.start()
    +
    +    taskId = taskInfo.getTaskId
    +  }
    +
    +  override def error(d: ExecutorDriver, message: String) {
    --- End diff --
    
    methods without return types should use `: Unit` (see https://cwiki.apache.org/confluence/display/SPARK/Spark+Code+Style+Guide)


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org


Mime
View raw message