diff options
author | zsxwing <zsxwing@gmail.com> | 2015-04-04 11:52:05 -0700 |
---|---|---|
committer | Reynold Xin <rxin@databricks.com> | 2015-04-04 11:52:05 -0700 |
commit | f15806a8f8ca34288ddb2d74b9ff1972c8374b59 (patch) | |
tree | 88abe5de9fadf078e57951450cb3368d0fb7cb64 /yarn/src | |
parent | 7bca62f79056e592cf07b49d8b8d04c59dea25fc (diff) | |
download | spark-f15806a8f8ca34288ddb2d74b9ff1972c8374b59.tar.gz spark-f15806a8f8ca34288ddb2d74b9ff1972c8374b59.tar.bz2 spark-f15806a8f8ca34288ddb2d74b9ff1972c8374b59.zip |
[SPARK-6602][Core] Replace direct use of Akka with Spark RPC interface - part 1
This PR replaced the following `Actor`s to `RpcEndpoint`:
1. HeartbeatReceiver
1. ExecutorActor
1. BlockManagerMasterActor
1. BlockManagerSlaveActor
1. CoarseGrainedExecutorBackend and subclasses
1. CoarseGrainedSchedulerBackend.DriverActor
This is the first PR. I will split the work of SPARK-6602 to several PRs for code review.
Author: zsxwing <zsxwing@gmail.com>
Closes #5268 from zsxwing/rpc-rewrite and squashes the following commits:
287e9f8 [zsxwing] Fix the code style
26c56b7 [zsxwing] Merge branch 'master' into rpc-rewrite
9cc825a [zsxwing] Rmove setupThreadSafeEndpoint and add ThreadSafeRpcEndpoint
30a9036 [zsxwing] Make self return null after stopping RpcEndpointRef; fix docs and error messages
705245d [zsxwing] Fix some bugs after rebasing the changes on the master
003cf80 [zsxwing] Update CoarseGrainedExecutorBackend and CoarseGrainedSchedulerBackend to use RpcEndpoint
7d0e6dc [zsxwing] Update BlockManagerSlaveActor to use RpcEndpoint
f5d6543 [zsxwing] Update BlockManagerMaster to use RpcEndpoint
30e3f9f [zsxwing] Update ExecutorActor to use RpcEndpoint
478b443 [zsxwing] Update HeartbeatReceiver to use RpcEndpoint
Diffstat (limited to 'yarn/src')
-rw-r--r-- | yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala | 86 | ||||
-rw-r--r-- | yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala | 2 |
2 files changed, 39 insertions, 49 deletions
diff --git a/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala b/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala index 455554eea0..24a1e02795 100644 --- a/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala +++ b/yarn/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala @@ -24,22 +24,20 @@ import java.lang.reflect.InvocationTargetException import java.net.{Socket, URL} import java.util.concurrent.atomic.AtomicReference -import akka.actor._ -import akka.remote._ import org.apache.hadoop.fs.{FileSystem, Path} import org.apache.hadoop.util.ShutdownHookManager import org.apache.hadoop.yarn.api._ import org.apache.hadoop.yarn.api.records._ import org.apache.hadoop.yarn.conf.YarnConfiguration +import org.apache.spark.rpc._ import org.apache.spark.{Logging, SecurityManager, SparkConf, SparkContext, SparkEnv} import org.apache.spark.SparkException import org.apache.spark.deploy.{PythonRunner, SparkHadoopUtil} import org.apache.spark.deploy.history.HistoryServer import org.apache.spark.scheduler.cluster.YarnSchedulerBackend import org.apache.spark.scheduler.cluster.CoarseGrainedClusterMessages._ -import org.apache.spark.util.{AkkaUtils, ChildFirstURLClassLoader, MutableURLClassLoader, - SignalLogger, Utils} +import org.apache.spark.util._ /** * Common application master functionality for Spark on Yarn. @@ -72,8 +70,8 @@ private[spark] class ApplicationMaster( @volatile private var allocator: YarnAllocator = _ // Fields used in client mode. - private var actorSystem: ActorSystem = null - private var actor: ActorRef = _ + private var rpcEnv: RpcEnv = null + private var amEndpoint: RpcEndpointRef = _ // Fields used in cluster mode. private val sparkContextRef = new AtomicReference[SparkContext](null) @@ -240,22 +238,21 @@ private[spark] class ApplicationMaster( } /** - * Create an actor that communicates with the driver. + * Create an [[RpcEndpoint]] that communicates with the driver. * * In cluster mode, the AM and the driver belong to same process - * so the AM actor need not monitor lifecycle of the driver. + * so the AMEndpoint need not monitor lifecycle of the driver. */ - private def runAMActor( + private def runAMEndpoint( host: String, port: String, isClusterMode: Boolean): Unit = { - val driverUrl = AkkaUtils.address( - AkkaUtils.protocol(actorSystem), + val driverEndpont = rpcEnv.setupEndpointRef( SparkEnv.driverActorSystemName, - host, - port, - YarnSchedulerBackend.ACTOR_NAME) - actor = actorSystem.actorOf(Props(new AMActor(driverUrl, isClusterMode)), name = "YarnAM") + RpcAddress(host, port.toInt), + YarnSchedulerBackend.ENDPOINT_NAME) + amEndpoint = + rpcEnv.setupEndpoint("YarnAM", new AMEndpoint(rpcEnv, driverEndpont, isClusterMode)) } private def runDriver(securityMgr: SecurityManager): Unit = { @@ -272,8 +269,8 @@ private[spark] class ApplicationMaster( ApplicationMaster.EXIT_SC_NOT_INITED, "Timed out waiting for SparkContext.") } else { - actorSystem = sc.env.actorSystem - runAMActor( + rpcEnv = sc.env.rpcEnv + runAMEndpoint( sc.getConf.get("spark.driver.host"), sc.getConf.get("spark.driver.port"), isClusterMode = true) @@ -283,8 +280,7 @@ private[spark] class ApplicationMaster( } private def runExecutorLauncher(securityMgr: SecurityManager): Unit = { - actorSystem = AkkaUtils.createActorSystem("sparkYarnAM", Utils.localHostName, 0, - conf = sparkConf, securityManager = securityMgr)._1 + rpcEnv = RpcEnv.create("sparkYarnAM", Utils.localHostName, 0, sparkConf, securityMgr) waitForSparkDriver() addAmIpFilter() registerAM(sparkConf.get("spark.driver.appUIAddress", ""), securityMgr) @@ -431,7 +427,7 @@ private[spark] class ApplicationMaster( sparkConf.set("spark.driver.host", driverHost) sparkConf.set("spark.driver.port", driverPort.toString) - runAMActor(driverHost, driverPort.toString, isClusterMode = false) + runAMEndpoint(driverHost, driverPort.toString, isClusterMode = false) } /** Add the Yarn IP filter that is required for properly securing the UI. */ @@ -443,7 +439,7 @@ private[spark] class ApplicationMaster( System.setProperty("spark.ui.filters", amFilter) params.foreach { case (k, v) => System.setProperty(s"spark.$amFilter.param.$k", v) } } else { - actor ! AddWebUIFilter(amFilter, params.toMap, proxyBase) + amEndpoint.send(AddWebUIFilter(amFilter, params.toMap, proxyBase)) } } @@ -505,44 +501,29 @@ private[spark] class ApplicationMaster( } /** - * An actor that communicates with the driver's scheduler backend. + * An [[RpcEndpoint]] that communicates with the driver's scheduler backend. */ - private class AMActor(driverUrl: String, isClusterMode: Boolean) extends Actor { - var driver: ActorSelection = _ - - override def preStart(): Unit = { - logInfo("Listen to driver: " + driverUrl) - driver = context.actorSelection(driverUrl) - // Send a hello message to establish the connection, after which - // we can monitor Lifecycle Events. - driver ! "Hello" - driver ! RegisterClusterManager - // In cluster mode, the AM can directly monitor the driver status instead - // of trying to deduce it from the lifecycle of the driver's actor - if (!isClusterMode) { - context.system.eventStream.subscribe(self, classOf[RemotingLifecycleEvent]) - } + private class AMEndpoint( + override val rpcEnv: RpcEnv, driver: RpcEndpointRef, isClusterMode: Boolean) + extends RpcEndpoint with Logging { + + override def onStart(): Unit = { + driver.send(RegisterClusterManager(self)) } override def receive: PartialFunction[Any, Unit] = { - case x: DisassociatedEvent => - logInfo(s"Driver terminated or disconnected! Shutting down. $x") - // In cluster mode, do not rely on the disassociated event to exit - // This avoids potentially reporting incorrect exit codes if the driver fails - if (!isClusterMode) { - finish(FinalApplicationStatus.SUCCEEDED, ApplicationMaster.EXIT_SUCCESS) - } - case x: AddWebUIFilter => logInfo(s"Add WebUI Filter. $x") - driver ! x + driver.send(x) + } + override def receiveAndReply(context: RpcCallContext): PartialFunction[Any, Unit] = { case RequestExecutors(requestedTotal) => Option(allocator) match { case Some(a) => a.requestTotalExecutors(requestedTotal) case None => logWarning("Container allocator is not ready to request executors yet.") } - sender ! true + context.reply(true) case KillExecutors(executorIds) => logInfo(s"Driver requested to kill executor(s) ${executorIds.mkString(", ")}.") @@ -550,7 +531,16 @@ private[spark] class ApplicationMaster( case Some(a) => executorIds.foreach(a.killExecutor) case None => logWarning("Container allocator is not ready to kill executors yet.") } - sender ! true + context.reply(true) + } + + override def onDisconnected(remoteAddress: RpcAddress): Unit = { + logInfo(s"Driver terminated or disconnected! Shutting down. $remoteAddress") + // In cluster mode, do not rely on the disassociated event to exit + // This avoids potentially reporting incorrect exit codes if the driver fails + if (!isClusterMode) { + finish(FinalApplicationStatus.SUCCEEDED, ApplicationMaster.EXIT_SUCCESS) + } } } diff --git a/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala b/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala index c98763e15b..b8f42dadcb 100644 --- a/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala +++ b/yarn/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala @@ -112,7 +112,7 @@ private[yarn] class YarnAllocator( SparkEnv.driverActorSystemName, sparkConf.get("spark.driver.host"), sparkConf.get("spark.driver.port"), - CoarseGrainedSchedulerBackend.ACTOR_NAME) + CoarseGrainedSchedulerBackend.ENDPOINT_NAME) // For testing private val launchContainers = sparkConf.getBoolean("spark.yarn.launchContainers", true) |