[SPARK-4857] [CORE] Adds Executor membership events to SparkListener

Adds onExecutorAdded and onExecutorRemoved events to the SparkListener. This will allow a client to get notified when an executor has been added/removed and provide additional information such as how many vcores it is consuming.

In addition, this commit adds a SparkListenerAdapter to the Java API that provides default implementations to the SparkListener. This is to get around the fact that default implementations for traits don't work in Java. Having Java clients extend SparkListenerAdapter moving forward will prevent breakage in java when we add new events to SparkListener.

Author: Kostas Sakellis <[email protected]>

Closes #3711 from ksakellis/kostas-spark-4857 and squashes the following commits:

946d2c5 [Kostas Sakellis] Added executorAdded/Removed events to MesosSchedulerBackend
b1d054a [Kostas Sakellis] Remove executorInfo from ExecutorRemoved event
1727b38 [Kostas Sakellis] Renamed ExecutorDetails back to ExecutorInfo and other CR feedback
14fe78d [Kostas Sakellis] Added executor added/removed events to json protocol
93d087b [Kostas Sakellis] [SPARK-4857] [CORE] Adds Executor membership events to SparkListener

Author: Kostas Sakellis

core/src/main/java/org/apache/spark/JavaSparkListener.java

@@ -0,0 +1,97 @@
+/*
+ * 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;
+
+import org.apache.spark.scheduler.SparkListener;
+import org.apache.spark.scheduler.SparkListenerApplicationEnd;
+import org.apache.spark.scheduler.SparkListenerApplicationStart;
+import org.apache.spark.scheduler.SparkListenerBlockManagerAdded;
+import org.apache.spark.scheduler.SparkListenerBlockManagerRemoved;
+import org.apache.spark.scheduler.SparkListenerEnvironmentUpdate;
+import org.apache.spark.scheduler.SparkListenerExecutorAdded;
+import org.apache.spark.scheduler.SparkListenerExecutorMetricsUpdate;
+import org.apache.spark.scheduler.SparkListenerExecutorRemoved;
+import org.apache.spark.scheduler.SparkListenerJobEnd;
+import org.apache.spark.scheduler.SparkListenerJobStart;
+import org.apache.spark.scheduler.SparkListenerStageCompleted;
+import org.apache.spark.scheduler.SparkListenerStageSubmitted;
+import org.apache.spark.scheduler.SparkListenerTaskEnd;
+import org.apache.spark.scheduler.SparkListenerTaskGettingResult;
+import org.apache.spark.scheduler.SparkListenerTaskStart;
+import org.apache.spark.scheduler.SparkListenerUnpersistRDD;
+
+/**
+ * Java clients should extend this class instead of implementing
+ * SparkListener directly. This is to prevent java clients
+ * from breaking when new events are added to the SparkListener
+ * trait.
+ *
+ * This is a concrete class instead of abstract to enforce
+ * new events get added to both the SparkListener and this adapter
+ * in lockstep.
+ */
+public class JavaSparkListener implements SparkListener {
+
+  @Override
+  public void onStageCompleted(SparkListenerStageCompleted stageCompleted) { }
+
+  @Override
+  public void onStageSubmitted(SparkListenerStageSubmitted stageSubmitted) { }
+
+  @Override
+  public void onTaskStart(SparkListenerTaskStart taskStart) { }
+
+  @Override
+  public void onTaskGettingResult(SparkListenerTaskGettingResult taskGettingResult) { }
+
+  @Override
+  public void onTaskEnd(SparkListenerTaskEnd taskEnd) { }
+
+  @Override
+  public void onJobStart(SparkListenerJobStart jobStart) { }
+
+  @Override
+  public void onJobEnd(SparkListenerJobEnd jobEnd) { }
+
+  @Override
+  public void onEnvironmentUpdate(SparkListenerEnvironmentUpdate environmentUpdate) { }
+
+  @Override
+  public void onBlockManagerAdded(SparkListenerBlockManagerAdded blockManagerAdded) { }
+
+  @Override
+  public void onBlockManagerRemoved(SparkListenerBlockManagerRemoved blockManagerRemoved) { }
+
+  @Override
+  public void onUnpersistRDD(SparkListenerUnpersistRDD unpersistRDD) { }
+
+  @Override
+  public void onApplicationStart(SparkListenerApplicationStart applicationStart) { }
+
+  @Override
+  public void onApplicationEnd(SparkListenerApplicationEnd applicationEnd) { }
+
+  @Override
+  public void onExecutorMetricsUpdate(SparkListenerExecutorMetricsUpdate executorMetricsUpdate) { }
+
+  @Override
+  public void onExecutorAdded(SparkListenerExecutorAdded executorAdded) { }
+
+  @Override
+  public void onExecutorRemoved(SparkListenerExecutorRemoved executorRemoved) { }
+}

core/src/main/scala/org/apache/spark/deploy/master/ApplicationInfo.scala

@@ -38,8 +38,8 @@ private[spark] class ApplicationInfo(
   extends Serializable {
 
   @transient var state: ApplicationState.Value = _
-  @transient var executors: mutable.HashMap[Int, ExecutorInfo] = _
-  @transient var removedExecutors: ArrayBuffer[ExecutorInfo] = _
+  @transient var executors: mutable.HashMap[Int, ExecutorDesc] = _
+  @transient var removedExecutors: ArrayBuffer[ExecutorDesc] = _
   @transient var coresGranted: Int = _
   @transient var endTime: Long = _
   @transient var appSource: ApplicationSource = _
@@ -55,12 +55,12 @@ private[spark] class ApplicationInfo(
 
   private def init() {
     state = ApplicationState.WAITING
-    executors = new mutable.HashMap[Int, ExecutorInfo]
+    executors = new mutable.HashMap[Int, ExecutorDesc]
     coresGranted = 0
     endTime = -1L
     appSource = new ApplicationSource(this)
     nextExecutorId = 0
-    removedExecutors = new ArrayBuffer[ExecutorInfo]
+    removedExecutors = new ArrayBuffer[ExecutorDesc]
   }
 
   private def newExecutorId(useID: Option[Int] = None): Int = {
@@ -75,14 +75,14 @@ private[spark] class ApplicationInfo(
     }
   }
 
-  def addExecutor(worker: WorkerInfo, cores: Int, useID: Option[Int] = None): ExecutorInfo = {
-    val exec = new ExecutorInfo(newExecutorId(useID), this, worker, cores, desc.memoryPerSlave)
+  def addExecutor(worker: WorkerInfo, cores: Int, useID: Option[Int] = None): ExecutorDesc = {
+    val exec = new ExecutorDesc(newExecutorId(useID), this, worker, cores, desc.memoryPerSlave)
     executors(exec.id) = exec
     coresGranted += cores
     exec
   }
 
-  def removeExecutor(exec: ExecutorInfo) {
+  def removeExecutor(exec: ExecutorDesc) {
     if (executors.contains(exec.id)) {
       removedExecutors += executors(exec.id)
       executors -= exec.id

core/src/main/scala/org/apache/spark/deploy/master/ExecutorInfo.scala renamed to core/src/main/scala/org/apache/spark/deploy/master/ExecutorDesc.scala

@@ -19,7 +19,7 @@ package org.apache.spark.deploy.master
 
 import org.apache.spark.deploy.{ExecutorDescription, ExecutorState}
 
-private[spark] class ExecutorInfo(
+private[spark] class ExecutorDesc(
     val id: Int,
     val application: ApplicationInfo,
     val worker: WorkerInfo,
@@ -37,7 +37,7 @@ private[spark] class ExecutorInfo(
 
   override def equals(other: Any): Boolean = {
     other match {
-      case info: ExecutorInfo =>
+      case info: ExecutorDesc =>
         fullId == info.fullId &&
         worker.id == info.worker.id &&
         cores == info.cores &&

core/src/main/scala/org/apache/spark/deploy/master/Master.scala

@@ -581,7 +581,7 @@ private[spark] class Master(
     }
   }
 
-  def launchExecutor(worker: WorkerInfo, exec: ExecutorInfo) {
+  def launchExecutor(worker: WorkerInfo, exec: ExecutorDesc) {
     logInfo("Launching executor " + exec.fullId + " on worker " + worker.id)
     worker.addExecutor(exec)
     worker.actor ! LaunchExecutor(masterUrl,

core/src/main/scala/org/apache/spark/deploy/master/WorkerInfo.scala

@@ -38,7 +38,7 @@ private[spark] class WorkerInfo(
   Utils.checkHost(host, "Expected hostname")
   assert (port > 0)
 
-  @transient var executors: mutable.HashMap[String, ExecutorInfo] = _ // executorId => info
+  @transient var executors: mutable.HashMap[String, ExecutorDesc] = _ // executorId => info
   @transient var drivers: mutable.HashMap[String, DriverInfo] = _ // driverId => info
   @transient var state: WorkerState.Value = _
   @transient var coresUsed: Int = _
@@ -70,13 +70,13 @@ private[spark] class WorkerInfo(
     host + ":" + port
   }
 
-  def addExecutor(exec: ExecutorInfo) {
+  def addExecutor(exec: ExecutorDesc) {
     executors(exec.fullId) = exec
     coresUsed += exec.cores
     memoryUsed += exec.memory
   }
 
-  def removeExecutor(exec: ExecutorInfo) {
+  def removeExecutor(exec: ExecutorDesc) {
     if (executors.contains(exec.fullId)) {
       executors -= exec.fullId
       coresUsed -= exec.cores

core/src/main/scala/org/apache/spark/deploy/master/ui/ApplicationPage.scala

@@ -27,7 +27,7 @@ import org.json4s.JValue
 
 import org.apache.spark.deploy.{ExecutorState, JsonProtocol}
 import org.apache.spark.deploy.DeployMessages.{MasterStateResponse, RequestMasterState}
-import org.apache.spark.deploy.master.ExecutorInfo
+import org.apache.spark.deploy.master.ExecutorDesc
 import org.apache.spark.ui.{UIUtils, WebUIPage}
 import org.apache.spark.util.Utils
 
@@ -109,7 +109,7 @@ private[spark] class ApplicationPage(parent: MasterWebUI) extends WebUIPage("app
     UIUtils.basicSparkPage(content, "Application: " + app.desc.name)
   }
 
-  private def executorRow(executor: ExecutorInfo): Seq[Node] = {
+  private def executorRow(executor: ExecutorDesc): Seq[Node] = {
     <tr>
       <td>{executor.id}</td>
       <td>

core/src/main/scala/org/apache/spark/scheduler/EventLoggingListener.scala

@@ -168,6 +168,10 @@ private[spark] class EventLoggingListener(
     logEvent(event, flushLogger = true)
   override def onApplicationEnd(event: SparkListenerApplicationEnd) =
     logEvent(event, flushLogger = true)
+  override def onExecutorAdded(event: SparkListenerExecutorAdded) =
+    logEvent(event, flushLogger = true)
+  override def onExecutorRemoved(event: SparkListenerExecutorRemoved) =
+    logEvent(event, flushLogger = true)
 
   // No-op because logging every update would be overkill
   override def onExecutorMetricsUpdate(event: SparkListenerExecutorMetricsUpdate) { }

core/src/main/scala/org/apache/spark/scheduler/SparkListener.scala

@@ -25,6 +25,7 @@ import scala.collection.mutable
 import org.apache.spark.{Logging, TaskEndReason}
 import org.apache.spark.annotation.DeveloperApi
 import org.apache.spark.executor.TaskMetrics
+import org.apache.spark.scheduler.cluster.ExecutorInfo
 import org.apache.spark.storage.BlockManagerId
 import org.apache.spark.util.{Distribution, Utils}
 
@@ -84,6 +85,14 @@ case class SparkListenerBlockManagerRemoved(time: Long, blockManagerId: BlockMan
 @DeveloperApi
 case class SparkListenerUnpersistRDD(rddId: Int) extends SparkListenerEvent
 
+@DeveloperApi
+case class SparkListenerExecutorAdded(executorId: String, executorInfo: ExecutorInfo)
+  extends SparkListenerEvent
+
+@DeveloperApi
+case class SparkListenerExecutorRemoved(executorId: String)
+  extends SparkListenerEvent
+
 /**
  * Periodic updates from executors.
  * @param execId executor id
@@ -109,7 +118,8 @@ private[spark] case object SparkListenerShutdown extends SparkListenerEvent
 /**
  * :: DeveloperApi ::
  * Interface for listening to events from the Spark scheduler. Note that this is an internal
- * interface which might change in different Spark releases.
+ * interface which might change in different Spark releases. Java clients should extend
+ * {@link JavaSparkListener}
  */
 @DeveloperApi
 trait SparkListener {
@@ -183,6 +193,16 @@ trait SparkListener {
    * Called when the driver receives task metrics from an executor in a heartbeat.
    */
   def onExecutorMetricsUpdate(executorMetricsUpdate: SparkListenerExecutorMetricsUpdate) { }
+
+  /**
+   * Called when the driver registers a new executor.
+   */
+  def onExecutorAdded(executorAdded: SparkListenerExecutorAdded) { }
+
+  /**
+   * Called when the driver removes an executor.
+   */
+  def onExecutorRemoved(executorRemoved: SparkListenerExecutorRemoved) { }
 }
 
 /**

core/src/main/scala/org/apache/spark/scheduler/SparkListenerBus.scala

@@ -70,6 +70,10 @@ private[spark] trait SparkListenerBus extends Logging {
         foreachListener(_.onApplicationEnd(applicationEnd))
       case metricsUpdate: SparkListenerExecutorMetricsUpdate =>
         foreachListener(_.onExecutorMetricsUpdate(metricsUpdate))
+      case executorAdded: SparkListenerExecutorAdded =>
+        foreachListener(_.onExecutorAdded(executorAdded))
+      case executorRemoved: SparkListenerExecutorRemoved =>
+        foreachListener(_.onExecutorRemoved(executorRemoved))
       case SparkListenerShutdown =>
     }
   }

core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedSchedulerBackend.scala

@@ -28,7 +28,7 @@ import akka.pattern.ask
 import akka.remote.{DisassociatedEvent, RemotingLifecycleEvent}
 
 import org.apache.spark.{ExecutorAllocationClient, Logging, SparkEnv, SparkException, TaskState}
-import org.apache.spark.scheduler.{SchedulerBackend, SlaveLost, TaskDescription, TaskSchedulerImpl, WorkerOffer}
+import org.apache.spark.scheduler._
 import org.apache.spark.scheduler.cluster.CoarseGrainedClusterMessages._
 import org.apache.spark.util.{ActorLogReceive, SerializableBuffer, AkkaUtils, Utils}
 
@@ -66,6 +66,8 @@ class CoarseGrainedSchedulerBackend(scheduler: TaskSchedulerImpl, val actorSyste
   // Number of executors requested from the cluster manager that have not registered yet
   private var numPendingExecutors = 0
 
+  private val listenerBus = scheduler.sc.listenerBus
+
   // Executors we have requested the cluster manager to kill that have not died yet
   private val executorsPendingToRemove = new HashSet[String]
 
@@ -106,6 +108,7 @@ class CoarseGrainedSchedulerBackend(scheduler: TaskSchedulerImpl, val actorSyste
               logDebug(s"Decremented number of pending executors ($numPendingExecutors left)")
             }
           }
+          listenerBus.post(SparkListenerExecutorAdded(executorId, data))
           makeOffers()
         }
 
@@ -213,6 +216,7 @@ class CoarseGrainedSchedulerBackend(scheduler: TaskSchedulerImpl, val actorSyste
           totalCoreCount.addAndGet(-executorInfo.totalCores)
           totalRegisteredExecutors.addAndGet(-1)
           scheduler.executorLost(executorId, SlaveLost(reason))
+          listenerBus.post(SparkListenerExecutorRemoved(executorId))
         case None => logError(s"Asked to remove non-existent executor $executorId")
       }
     }