stop SparkContext instead of quit the JVM process

Author: CodingCat

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

@@ -145,7 +145,7 @@ private[spark] class ContextCleaner(sc: SparkContext) extends Logging {
   }
 
   /** Keep cleaning RDD, shuffle, and broadcast state. */
-  private def keepCleaning(): Unit = Utils.tryOrExit {
+  private def keepCleaning(): Unit = Utils.tryOrStopSparkContext(sc) {
     while (!stopped) {
       try {
         val reference = Option(referenceQueue.remove(ContextCleaner.REF_QUEUE_POLL_TIMEOUT))

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

@@ -1736,7 +1736,7 @@ class SparkContext(config: SparkConf) extends Logging with ExecutorAllocationCli
         }
     }
 
-    listenerBus.start()
+    listenerBus.start(this)
   }
 
   /** Post the application start event */

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

@@ -145,7 +145,7 @@ private[spark] class TaskSchedulerImpl(
       import sc.env.actorSystem.dispatcher
       sc.env.actorSystem.scheduler.schedule(SPECULATION_INTERVAL milliseconds,
             SPECULATION_INTERVAL milliseconds) {
-        Utils.tryOrExit { checkSpeculatableTasks() }
+        Utils.tryOrStopSparkContext(sc) { checkSpeculatableTasks() }
       }
     }
   }

core/src/main/scala/org/apache/spark/util/AsynchronousListenerBus.scala

@@ -21,6 +21,7 @@ import java.util.concurrent._
 import java.util.concurrent.atomic.AtomicBoolean
 
 import com.google.common.annotations.VisibleForTesting
+import org.apache.spark.SparkContext
 
 /**
  * Asynchronously passes events to registered listeners.
@@ -38,6 +39,8 @@ private[spark] abstract class AsynchronousListenerBus[L <: AnyRef, E](name: Stri
 
   self =>
 
+  private var sparkContext: SparkContext = null
+  
   /* Cap the capacity of the event queue so we get an explicit error (rather than
    * an OOM exception) if it's perpetually being added to more quickly than it's being drained. */
   private val EVENT_QUEUE_CAPACITY = 10000
@@ -57,7 +60,7 @@ private[spark] abstract class AsynchronousListenerBus[L <: AnyRef, E](name: Stri
 
   private val listenerThread = new Thread(name) {
     setDaemon(true)
-    override def run(): Unit = Utils.tryOrExit {
+    override def run(): Unit = Utils.tryOrStopSparkContext(sparkContext) {
       while (true) {
         eventLock.acquire()
         self.synchronized {
@@ -90,8 +93,9 @@ private[spark] abstract class AsynchronousListenerBus[L <: AnyRef, E](name: Stri
    * listens for any additional events asynchronously while the listener bus is still running.
    * This should only be called once.
    */
-  def start() {
+  def start(sc: SparkContext) {
     if (started.compareAndSet(false, true)) {
+      sparkContext = sc
       listenerThread.start()
     } else {
       throw new IllegalStateException(s"$name already started!")

core/src/main/scala/org/apache/spark/util/Utils.scala

@@ -1155,6 +1155,19 @@ private[spark] object Utils extends Logging {
     }
   }
 
+  /**
+   * Execute a block of code that evaluates to Unit, forwarding any uncaught exceptions to the
+   * default UncaughtExceptionHandler
+   */
+  def tryOrStopSparkContext(sc: SparkContext)(block: => Unit) {
+    try {
+      block
+    } catch {
+      case e: ControlThrowable => throw e
+      case t: Throwable => sc.stop()
+    }
+  }
+
   /**
    * Execute a block of code that evaluates to Unit, re-throwing any non-fatal uncaught
    * exceptions as IOException.  This is used when implementing Externalizable and Serializable's

core/src/test/scala/org/apache/spark/scheduler/EventLoggingListenerSuite.scala

@@ -25,9 +25,9 @@ import scala.io.Source
 
 import org.apache.hadoop.fs.Path
 import org.json4s.jackson.JsonMethods._
-import org.scalatest.{BeforeAndAfter, FunSuite}
+import org.scalatest.{FunSuiteLike, BeforeAndAfter, FunSuite}
 
-import org.apache.spark.{Logging, SparkConf, SparkContext, SPARK_VERSION}
+import org.apache.spark._
 import org.apache.spark.deploy.SparkHadoopUtil
 import org.apache.spark.io._
 import org.apache.spark.util.{JsonProtocol, Utils}
@@ -39,7 +39,8 @@ import org.apache.spark.util.{JsonProtocol, Utils}
  * logging events, whether the parsing of the file names is correct, and whether the logged events
  * can be read and deserialized into actual SparkListenerEvents.
  */
-class EventLoggingListenerSuite extends FunSuite with BeforeAndAfter with Logging {
+class EventLoggingListenerSuite extends FunSuite with LocalSparkContext with BeforeAndAfter
+  with Logging {
   import EventLoggingListenerSuite._
 
   private val fileSystem = Utils.getHadoopFileSystem("/",
@@ -144,7 +145,7 @@ class EventLoggingListenerSuite extends FunSuite with BeforeAndAfter with Loggin
 
     // A comprehensive test on JSON de/serialization of all events is in JsonProtocolSuite
     eventLogger.start()
-    listenerBus.start()
+    listenerBus.start(sc)
     listenerBus.addListener(eventLogger)
     listenerBus.postToAll(applicationStart)
     listenerBus.postToAll(applicationEnd)

core/src/test/scala/org/apache/spark/scheduler/SparkListenerSuite.scala

@@ -46,7 +46,7 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers
     assert(counter.count === 0)
 
     // Starting listener bus should flush all buffered events
-    bus.start()
+    bus.start(sc)
     assert(bus.waitUntilEmpty(WAIT_TIMEOUT_MILLIS))
     assert(counter.count === 5)
 
@@ -58,8 +58,8 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers
     // Listener bus must not be started twice
     intercept[IllegalStateException] {
       val bus = new LiveListenerBus
-      bus.start()
-      bus.start()
+      bus.start(sc)
+      bus.start(sc)
     }
 
     // ... or stopped before starting
@@ -96,7 +96,7 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers
     val blockingListener = new BlockingListener
 
     bus.addListener(blockingListener)
-    bus.start()
+    bus.start(sc)
     bus.post(SparkListenerJobEnd(0, jobCompletionTime, JobSucceeded))
 
     listenerStarted.acquire()
@@ -347,7 +347,7 @@ class SparkListenerSuite extends FunSuite with LocalSparkContext with Matchers
     bus.addListener(badListener)
     bus.addListener(jobCounter1)
     bus.addListener(jobCounter2)
-    bus.start()
+    bus.start(sc)
 
     // Post events to all listeners, and wait until the queue is drained
     (1 to 5).foreach { _ => bus.post(SparkListenerJobEnd(0, jobCompletionTime, JobSucceeded)) }

streaming/src/main/scala/org/apache/spark/streaming/scheduler/JobScheduler.scala

@@ -61,7 +61,7 @@ class JobScheduler(val ssc: StreamingContext) extends Logging {
       }
     }), "JobScheduler")
 
-    listenerBus.start()
+    listenerBus.start(ssc.sparkContext)
     receiverTracker = new ReceiverTracker(ssc)
     receiverTracker.start()
     jobGenerator.start()