Merge branch 'master' of github.com:apache/spark into binary

Author: Davies Liu

core/src/main/scala/org/apache/spark/SparkContext.scala

@@ -21,9 +21,8 @@ import scala.language.implicitConversions
 
 import java.io._
 import java.net.URI
-import java.util.Arrays
+import java.util.{Arrays, Properties, UUID}
 import java.util.concurrent.atomic.AtomicInteger
-import java.util.{Properties, UUID}
 import java.util.UUID.randomUUID
 import scala.collection.{Map, Set}
 import scala.collection.generic.Growable
@@ -41,6 +40,7 @@ import akka.actor.Props
 import org.apache.spark.annotation.{DeveloperApi, Experimental}
 import org.apache.spark.broadcast.Broadcast
 import org.apache.spark.deploy.{LocalSparkCluster, SparkHadoopUtil}
+import org.apache.spark.executor.TriggerThreadDump
 import org.apache.spark.input.{StreamInputFormat, PortableDataStream, WholeTextFileInputFormat, FixedLengthBinaryInputFormat}
 import org.apache.spark.partial.{ApproximateEvaluator, PartialResult}
 import org.apache.spark.rdd._
@@ -51,7 +51,7 @@ import org.apache.spark.scheduler.local.LocalBackend
 import org.apache.spark.storage._
 import org.apache.spark.ui.SparkUI
 import org.apache.spark.ui.jobs.JobProgressListener
-import org.apache.spark.util.{CallSite, ClosureCleaner, MetadataCleaner, MetadataCleanerType, TimeStampedWeakValueHashMap, Utils}
+import org.apache.spark.util._
 
 /**
  * Main entry point for Spark functionality. A SparkContext represents the connection to a Spark
@@ -361,6 +361,29 @@ class SparkContext(config: SparkConf) extends SparkStatusAPI with Logging {
     override protected def childValue(parent: Properties): Properties = new Properties(parent)
   }
 
+  /**
+   * Called by the web UI to obtain executor thread dumps.  This method may be expensive.
+   * Logs an error and returns None if we failed to obtain a thread dump, which could occur due
+   * to an executor being dead or unresponsive or due to network issues while sending the thread
+   * dump message back to the driver.
+   */
+  private[spark] def getExecutorThreadDump(executorId: String): Option[Array[ThreadStackTrace]] = {
+    try {
+      if (executorId == SparkContext.DRIVER_IDENTIFIER) {
+        Some(Utils.getThreadDump())
+      } else {
+        val (host, port) = env.blockManager.master.getActorSystemHostPortForExecutor(executorId).get
+        val actorRef = AkkaUtils.makeExecutorRef("ExecutorActor", conf, host, port, env.actorSystem)
+        Some(AkkaUtils.askWithReply[Array[ThreadStackTrace]](TriggerThreadDump, actorRef,
+          AkkaUtils.numRetries(conf), AkkaUtils.retryWaitMs(conf), AkkaUtils.askTimeout(conf)))
+      }
+    } catch {
+      case e: Exception =>
+        logError(s"Exception getting thread dump from executor $executorId", e)
+        None
+    }
+  }
+
   private[spark] def getLocalProperties: Properties = localProperties.get()
 
   private[spark] def setLocalProperties(props: Properties) {

core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala

@@ -131,7 +131,8 @@ private[spark] object CoarseGrainedExecutorBackend extends Logging {
       // Create a new ActorSystem using driver's Spark properties to run the backend.
       val driverConf = new SparkConf().setAll(props)
       val (actorSystem, boundPort) = AkkaUtils.createActorSystem(
-        "sparkExecutor", hostname, port, driverConf, new SecurityManager(driverConf))
+        SparkEnv.executorActorSystemName,
+        hostname, port, driverConf, new SecurityManager(driverConf))
       // set it
       val sparkHostPort = hostname + ":" + boundPort
       actorSystem.actorOf(

core/src/main/scala/org/apache/spark/executor/Executor.scala

@@ -26,7 +26,7 @@ import scala.collection.JavaConversions._
 import scala.collection.mutable.{ArrayBuffer, HashMap}
 import scala.util.control.NonFatal
 
-import akka.actor.ActorSystem
+import akka.actor.{Props, ActorSystem}
 
 import org.apache.spark._
 import org.apache.spark.deploy.SparkHadoopUtil
@@ -92,6 +92,10 @@ private[spark] class Executor(
     }
   }
 
+  // Create an actor for receiving RPCs from the driver
+  private val executorActor = env.actorSystem.actorOf(
+    Props(new ExecutorActor(executorId)), "ExecutorActor")
+
   // Create our ClassLoader
   // do this after SparkEnv creation so can access the SecurityManager
   private val urlClassLoader = createClassLoader()
@@ -131,6 +135,7 @@ private[spark] class Executor(
 
   def stop() {
     env.metricsSystem.report()
+    env.actorSystem.stop(executorActor)
     isStopped = true
     threadPool.shutdown()
     if (!isLocal) {

core/src/main/scala/org/apache/spark/executor/ExecutorActor.scala

@@ -0,0 +1,41 @@
+/*
+ * 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 akka.actor.Actor
+import org.apache.spark.Logging
+
+import org.apache.spark.util.{Utils, ActorLogReceive}
+
+/**
+ * Driver -> Executor message to trigger a thread dump.
+ */
+private[spark] case object TriggerThreadDump
+
+/**
+ * Actor that runs inside of executors to enable driver -> executor RPC.
+ */
+private[spark]
+class ExecutorActor(executorId: String) extends Actor with ActorLogReceive with Logging {
+
+  override def receiveWithLogging = {
+    case TriggerThreadDump =>
+      sender ! Utils.getThreadDump()
+  }
+
+}

core/src/main/scala/org/apache/spark/storage/BlockManagerMaster.scala

@@ -88,6 +88,10 @@ class BlockManagerMaster(
     askDriverWithReply[Seq[BlockManagerId]](GetPeers(blockManagerId))
   }
 
+  def getActorSystemHostPortForExecutor(executorId: String): Option[(String, Int)] = {
+    askDriverWithReply[Option[(String, Int)]](GetActorSystemHostPortForExecutor(executorId))
+  }
+
   /**
    * Remove a block from the slaves that have it. This can only be used to remove
    * blocks that the driver knows about.

core/src/main/scala/org/apache/spark/storage/BlockManagerMasterActor.scala

@@ -86,6 +86,9 @@ class BlockManagerMasterActor(val isLocal: Boolean, conf: SparkConf, listenerBus
     case GetPeers(blockManagerId) =>
       sender ! getPeers(blockManagerId)
 
+    case GetActorSystemHostPortForExecutor(executorId) =>
+      sender ! getActorSystemHostPortForExecutor(executorId)
+
     case GetMemoryStatus =>
       sender ! memoryStatus
 
@@ -412,6 +415,21 @@ class BlockManagerMasterActor(val isLocal: Boolean, conf: SparkConf, listenerBus
       Seq.empty
     }
   }
+
+  /**
+   * Returns the hostname and port of an executor's actor system, based on the Akka address of its
+   * BlockManagerSlaveActor.
+   */
+  private def getActorSystemHostPortForExecutor(executorId: String): Option[(String, Int)] = {
+    for (
+      blockManagerId <- blockManagerIdByExecutor.get(executorId);
+      info <- blockManagerInfo.get(blockManagerId);
+      host <- info.slaveActor.path.address.host;
+      port <- info.slaveActor.path.address.port
+    ) yield {
+      (host, port)
+    }
+  }
 }
 
 @DeveloperApi

core/src/main/scala/org/apache/spark/storage/BlockManagerMessages.scala

@@ -92,6 +92,8 @@ private[spark] object BlockManagerMessages {
 
   case class GetPeers(blockManagerId: BlockManagerId) extends ToBlockManagerMaster
 
+  case class GetActorSystemHostPortForExecutor(executorId: String) extends ToBlockManagerMaster
+
   case class RemoveExecutor(execId: String) extends ToBlockManagerMaster
 
   case object StopBlockManagerMaster extends ToBlockManagerMaster

core/src/main/scala/org/apache/spark/ui/exec/ExecutorThreadDumpPage.scala

@@ -0,0 +1,73 @@
+/*
+ * 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.ui.exec
+
+import javax.servlet.http.HttpServletRequest
+
+import scala.util.Try
+import scala.xml.{Text, Node}
+
+import org.apache.spark.ui.{UIUtils, WebUIPage}
+
+private[ui] class ExecutorThreadDumpPage(parent: ExecutorsTab) extends WebUIPage("threadDump") {
+
+  private val sc = parent.sc
+
+  def render(request: HttpServletRequest): Seq[Node] = {
+    val executorId = Option(request.getParameter("executorId")).getOrElse {
+      return Text(s"Missing executorId parameter")
+    }
+    val time = System.currentTimeMillis()
+    val maybeThreadDump = sc.get.getExecutorThreadDump(executorId)
+
+    val content = maybeThreadDump.map { threadDump =>
+      val dumpRows = threadDump.map { thread =>
+        <div class="accordion-group">
+          <div class="accordion-heading" onclick="$(this).next().toggleClass('hidden')">
+            <a class="accordion-toggle">
+              Thread {thread.threadId}: {thread.threadName} ({thread.threadState})
+            </a>
+          </div>
+          <div class="accordion-body hidden">
+            <div class="accordion-inner">
+              <pre>{thread.stackTrace}</pre>
+            </div>
+          </div>
+        </div>
+      }
+
+      <div class="row-fluid">
+        <p>Updated at {UIUtils.formatDate(time)}</p>
+        {
+          // scalastyle:off
+          <p><a class="expandbutton"
+                onClick="$('.accordion-body').removeClass('hidden'); $('.expandbutton').toggleClass('hidden')">
+            Expand All
+          </a></p>
+          <p><a class="expandbutton hidden"
+                onClick="$('.accordion-body').addClass('hidden'); $('.expandbutton').toggleClass('hidden')">
+            Collapse All
+          </a></p>
+          // scalastyle:on
+        }
+        <div class="accordion">{dumpRows}</div>
+      </div>
+    }.getOrElse(Text("Error fetching thread dump"))
+    UIUtils.headerSparkPage(s"Thread dump for executor $executorId", content, parent)
+  }
+}

core/src/main/scala/org/apache/spark/ui/exec/ExecutorsPage.scala

@@ -41,7 +41,10 @@ private case class ExecutorSummaryInfo(
     totalShuffleWrite: Long,
     maxMemory: Long)
 
-private[ui] class ExecutorsPage(parent: ExecutorsTab) extends WebUIPage("") {
+private[ui] class ExecutorsPage(
+    parent: ExecutorsTab,
+    threadDumpEnabled: Boolean)
+  extends WebUIPage("") {
   private val listener = parent.listener
 
   def render(request: HttpServletRequest): Seq[Node] = {
@@ -75,6 +78,7 @@ private[ui] class ExecutorsPage(parent: ExecutorsTab) extends WebUIPage("") {
               Shuffle Write
             </span>
           </th>
+          {if (threadDumpEnabled) <th class="sorttable_nosort">Thread Dump</th> else Seq.empty}
         </thead>
         <tbody>
           {execInfoSorted.map(execRow)}
@@ -133,6 +137,15 @@ private[ui] class ExecutorsPage(parent: ExecutorsTab) extends WebUIPage("") {
       <td sorttable_customkey={info.totalShuffleWrite.toString}>
         {Utils.bytesToString(info.totalShuffleWrite)}
       </td>
+      {
+        if (threadDumpEnabled) {
+          <td>
+            <a href={s"threadDump/?executorId=${info.id}"}>Thread Dump</a>
+          </td>
+        } else {
+          Seq.empty
+        }
+      }
     </tr>
   }
 

core/src/main/scala/org/apache/spark/ui/exec/ExecutorsTab.scala

@@ -27,8 +27,14 @@ import org.apache.spark.ui.{SparkUI, SparkUITab}
 
 private[ui] class ExecutorsTab(parent: SparkUI) extends SparkUITab(parent, "executors") {
   val listener = parent.executorsListener
+  val sc = parent.sc
+  val threadDumpEnabled =
+    sc.isDefined && parent.conf.getBoolean("spark.ui.threadDumpsEnabled", true)
 
-  attachPage(new ExecutorsPage(this))
+  attachPage(new ExecutorsPage(this, threadDumpEnabled))
+  if (threadDumpEnabled) {
+    attachPage(new ExecutorThreadDumpPage(this))
+  }
 }
 
 /**