[SPARK-2933] [yarn] Refactor and cleanup Yarn AM code.

This change modifies the Yarn module so that all the logic related
to running the ApplicationMaster is localized. Instead of, previously,
4 different classes with mostly identical code, now we have:

- A single, shared ApplicationMaster class, which can operate both in
  client and cluster mode, and substitutes the old ApplicationMaster
  (for cluster mode) and ExecutorLauncher (for client mode).

The benefit here is that all different execution modes for all supported
yarn versions use the same shared code for monitoring executor allocation,
setting up configuration, and monitoring the process's lifecycle.

- A new YarnRMClient interface, which defines basic RM functionality needed
  by the ApplicationMaster. This interface has concrete implementations for
  each supported Yarn version.

- A new YarnAllocator interface, which just abstracts the existing interface
  of the YarnAllocationHandler class. This is to avoid having to touch the
  allocator code too much in this change, although it might benefit from a
  similar effort in the future.

The end result is much easier to understand code, with much less duplication,
making it much easier to fix bugs, add features, and test everything knowing
that all supported versions will behave the same.

Author: Marcelo Vanzin <vanzin@cloudera.com>

Closes #2020 from vanzin/SPARK-2933 and squashes the following commits:

3bbf3e7 [Marcelo Vanzin] Merge branch 'master' into SPARK-2933
ff389ed [Marcelo Vanzin] Do not interrupt reporter thread from within itself.
3a8ed37 [Marcelo Vanzin] Remote stale comment.
0f5142c [Marcelo Vanzin] Review feedback.
41f8c8a [Marcelo Vanzin] Fix app status reporting.
c0794be [Marcelo Vanzin] Correctly clean up staging directory.
92770cc [Marcelo Vanzin] Merge branch 'master' into SPARK-2933
ecaf332 [Marcelo Vanzin] Small fix to shutdown code.
f02d3f8 [Marcelo Vanzin] Merge branch 'master' into SPARK-2933
f581122 [Marcelo Vanzin] Review feedback.
557fdeb [Marcelo Vanzin] Cleanup a couple more constants.
be6068d [Marcelo Vanzin] Restore shutdown hook to clean up staging dir.
5150993 [Marcelo Vanzin] Some more cleanup.
b6289ab [Marcelo Vanzin] Move cluster/client code to separate methods.
ecb23cd [Marcelo Vanzin] More trivial cleanup.
34f1e63 [Marcelo Vanzin] Fix some questionable error handling.
5657c7d [Marcelo Vanzin] Finish app if SparkContext initialization times out.
0e4be3d [Marcelo Vanzin] Keep "ExecutorLauncher" as the main class for client-mode AM.
91beabb [Marcelo Vanzin] Fix UI filter registration.
8c72239 [Marcelo Vanzin] Trivial cleanups.
99a52d5 [Marcelo Vanzin] Changes to the yarn-alpha project to use common AM code.
848ca6d [Marcelo Vanzin] [SPARK-2933] [yarn] Refactor and cleanup Yarn AM code.

10 files changed, 644 insertions, 62 deletions

diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala
new file mode 100644
index 0000000000..f15c813b83
--- /dev/null
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMaster.scala
@@ -0,0 +1,430 @@
+/*
+ * 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.deploy.yarn
+
+import java.io.IOException
+import java.net.Socket
+import java.util.concurrent.atomic.AtomicReference
+
+import scala.collection.JavaConversions._
+import scala.util.Try
+
+import akka.actor._
+import akka.remote._
+import org.apache.hadoop.conf.Configuration
+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.{Logging, SecurityManager, SparkConf, SparkContext, SparkEnv}
+import org.apache.spark.deploy.SparkHadoopUtil
+import org.apache.spark.scheduler.cluster.CoarseGrainedSchedulerBackend
+import org.apache.spark.scheduler.cluster.CoarseGrainedClusterMessages.AddWebUIFilter
+import org.apache.spark.util.{AkkaUtils, SignalLogger, Utils}
+
+/**
+ * Common application master functionality for Spark on Yarn.
+ */
+private[spark] class ApplicationMaster(args: ApplicationMasterArguments,
+  client: YarnRMClient) extends Logging {
+  // TODO: Currently, task to container is computed once (TaskSetManager) - which need not be
+  // optimal as more containers are available. Might need to handle this better.
+  private val ALLOCATE_HEARTBEAT_INTERVAL = 100
+
+  private val sparkConf = new SparkConf()
+  private val yarnConf: YarnConfiguration = new YarnConfiguration(new Configuration())
+  private val isDriver = args.userClass != null
+
+  // Default to numExecutors * 2, with minimum of 3
+  private val maxNumExecutorFailures = sparkConf.getInt("spark.yarn.max.executor.failures",
+    sparkConf.getInt("spark.yarn.max.worker.failures", math.max(args.numExecutors * 2, 3)))
+
+  @volatile private var finished = false
+  @volatile private var finalStatus = FinalApplicationStatus.UNDEFINED
+
+  private var reporterThread: Thread = _
+  private var allocator: YarnAllocator = _
+
+  // Fields used in client mode.
+  private var actorSystem: ActorSystem = null
+  private var actor: ActorRef = _
+
+  // Fields used in cluster mode.
+  private val sparkContextRef = new AtomicReference[SparkContext](null)
+
+  final def run(): Int = {
+    if (isDriver) {
+      // Set the web ui port to be ephemeral for yarn so we don't conflict with
+      // other spark processes running on the same box
+      System.setProperty("spark.ui.port", "0")
+
+      // Set the master property to match the requested mode.
+      System.setProperty("spark.master", "yarn-cluster")
+    }
+
+    logInfo("ApplicationAttemptId: " + client.getAttemptId())
+
+    val cleanupHook = new Runnable {
+      override def run() {
+        // If the SparkContext is still registered, shut it down as a best case effort in case
+        // users do not call sc.stop or do System.exit().
+        val sc = sparkContextRef.get()
+        if (sc != null) {
+          logInfo("Invoking sc stop from shutdown hook")
+          sc.stop()
+          finish(FinalApplicationStatus.SUCCEEDED)
+        }
+
+        // Cleanup the staging dir after the app is finished, or if it's the last attempt at
+        // running the AM.
+        val maxAppAttempts = client.getMaxRegAttempts(yarnConf)
+        val isLastAttempt = client.getAttemptId().getAttemptId() >= maxAppAttempts
+        if (finished || isLastAttempt) {
+          cleanupStagingDir()
+        }
+      }
+    }
+    // Use priority 30 as it's higher than HDFS. It's the same priority MapReduce is using.
+    ShutdownHookManager.get().addShutdownHook(cleanupHook, 30)
+
+    // Call this to force generation of secret so it gets populated into the
+    // Hadoop UGI. This has to happen before the startUserClass which does a
+    // doAs in order for the credentials to be passed on to the executor containers.
+    val securityMgr = new SecurityManager(sparkConf)
+
+    if (isDriver) {
+      runDriver()
+    } else {
+      runExecutorLauncher(securityMgr)
+    }
+
+    if (finalStatus != FinalApplicationStatus.UNDEFINED) {
+      finish(finalStatus)
+      0
+    } else {
+      1
+    }
+  }
+
+  final def finish(status: FinalApplicationStatus, diagnostics: String = null) = synchronized {
+    if (!finished) {
+      logInfo(s"Finishing ApplicationMaster with $status"  +
+        Option(diagnostics).map(msg => s" (diag message: $msg)").getOrElse(""))
+      finished = true
+      finalStatus = status
+      try {
+        if (Thread.currentThread() != reporterThread) {
+          reporterThread.interrupt()
+          reporterThread.join()
+        }
+      } finally {
+        client.shutdown(status, Option(diagnostics).getOrElse(""))
+      }
+    }
+  }
+
+  private def sparkContextInitialized(sc: SparkContext) = {
+    sparkContextRef.synchronized {
+      sparkContextRef.compareAndSet(null, sc)
+      sparkContextRef.notifyAll()
+    }
+  }
+
+  private def sparkContextStopped(sc: SparkContext) = {
+    sparkContextRef.compareAndSet(sc, null)
+  }
+
+  private def registerAM(uiAddress: String, uiHistoryAddress: String) = {
+    val sc = sparkContextRef.get()
+    allocator = client.register(yarnConf,
+      if (sc != null) sc.getConf else sparkConf,
+      if (sc != null) sc.preferredNodeLocationData else Map(),
+      uiAddress,
+      uiHistoryAddress)
+
+    allocator.allocateResources()
+    reporterThread = launchReporterThread()
+  }
+
+  private def runDriver(): Unit = {
+    addAmIpFilter()
+    val userThread = startUserClass()
+
+    // This a bit hacky, but we need to wait until the spark.driver.port property has
+    // been set by the Thread executing the user class.
+    val sc = waitForSparkContextInitialized()
+
+    // If there is no SparkContext at this point, just fail the app.
+    if (sc == null) {
+      finish(FinalApplicationStatus.FAILED, "Timed out waiting for SparkContext.")
+    } else {
+      registerAM(sc.ui.appUIHostPort, YarnSparkHadoopUtil.getUIHistoryAddress(sc, sparkConf))
+      try {
+        userThread.join()
+      } finally {
+        // In cluster mode, ask the reporter thread to stop since the user app is finished.
+        reporterThread.interrupt()
+      }
+    }
+  }
+
+  private def runExecutorLauncher(securityMgr: SecurityManager): Unit = {
+    actorSystem = AkkaUtils.createActorSystem("sparkYarnAM", Utils.localHostName, 0,
+      conf = sparkConf, securityManager = securityMgr)._1
+    actor = waitForSparkDriver()
+    addAmIpFilter()
+    registerAM(sparkConf.get("spark.driver.appUIAddress", ""),
+      sparkConf.get("spark.driver.appUIHistoryAddress", ""))
+
+    // In client mode the actor will stop the reporter thread.
+    reporterThread.join()
+    finalStatus = FinalApplicationStatus.SUCCEEDED
+  }
+
+  private def launchReporterThread(): Thread = {
+    // Ensure that progress is sent before YarnConfiguration.RM_AM_EXPIRY_INTERVAL_MS elapses.
+    val expiryInterval = yarnConf.getInt(YarnConfiguration.RM_AM_EXPIRY_INTERVAL_MS, 120000)
+
+    // we want to be reasonably responsive without causing too many requests to RM.
+    val schedulerInterval =
+      sparkConf.getLong("spark.yarn.scheduler.heartbeat.interval-ms", 5000)
+
+    // must be <= expiryInterval / 2.
+    val interval = math.max(0, math.min(expiryInterval / 2, schedulerInterval))
+
+    val t = new Thread {
+      override def run() {
+        while (!finished) {
+          checkNumExecutorsFailed()
+          if (!finished) {
+            logDebug("Sending progress")
+            allocator.allocateResources()
+            try {
+              Thread.sleep(interval)
+            } catch {
+              case e: InterruptedException =>
+            }
+          }
+        }
+      }
+    }
+    // setting to daemon status, though this is usually not a good idea.
+    t.setDaemon(true)
+    t.setName("Reporter")
+    t.start()
+    logInfo("Started progress reporter thread - sleep time : " + interval)
+    t
+  }
+
+  /**
+   * Clean up the staging directory.
+   */
+  private def cleanupStagingDir() {
+    val fs = FileSystem.get(yarnConf)
+    var stagingDirPath: Path = null
+    try {
+      val preserveFiles = sparkConf.get("spark.yarn.preserve.staging.files", "false").toBoolean
+      if (!preserveFiles) {
+        stagingDirPath = new Path(System.getenv("SPARK_YARN_STAGING_DIR"))
+        if (stagingDirPath == null) {
+          logError("Staging directory is null")
+          return
+        }
+        logInfo("Deleting staging directory " + stagingDirPath)
+        fs.delete(stagingDirPath, true)
+      }
+    } catch {
+      case ioe: IOException =>
+        logError("Failed to cleanup staging dir " + stagingDirPath, ioe)
+    }
+  }
+
+  private def waitForSparkContextInitialized(): SparkContext = {
+    logInfo("Waiting for spark context initialization")
+    try {
+      sparkContextRef.synchronized {
+        var count = 0
+        val waitTime = 10000L
+        val numTries = sparkConf.getInt("spark.yarn.ApplicationMaster.waitTries", 10)
+        while (sparkContextRef.get() == null && count < numTries && !finished) {
+          logInfo("Waiting for spark context initialization ... " + count)
+          count = count + 1
+          sparkContextRef.wait(waitTime)
+        }
+
+        val sparkContext = sparkContextRef.get()
+        assert(sparkContext != null || count >= numTries)
+        if (sparkContext == null) {
+          logError(
+            "Unable to retrieve sparkContext inspite of waiting for %d, numTries = %d".format(
+              count * waitTime, numTries))
+        }
+        sparkContext
+      }
+    }
+  }
+
+  private def waitForSparkDriver(): ActorRef = {
+    logInfo("Waiting for Spark driver to be reachable.")
+    var driverUp = false
+    val hostport = args.userArgs(0)
+    val (driverHost, driverPort) = Utils.parseHostPort(hostport)
+    while (!driverUp) {
+      try {
+        val socket = new Socket(driverHost, driverPort)
+        socket.close()
+        logInfo("Driver now available: %s:%s".format(driverHost, driverPort))
+        driverUp = true
+      } catch {
+        case e: Exception =>
+          logError("Failed to connect to driver at %s:%s, retrying ...".
+            format(driverHost, driverPort))
+          Thread.sleep(100)
+      }
+    }
+    sparkConf.set("spark.driver.host", driverHost)
+    sparkConf.set("spark.driver.port", driverPort.toString)
+
+    val driverUrl = "akka.tcp://%s@%s:%s/user/%s".format(
+      SparkEnv.driverActorSystemName,
+      driverHost,
+      driverPort.toString,
+      CoarseGrainedSchedulerBackend.ACTOR_NAME)
+    actorSystem.actorOf(Props(new MonitorActor(driverUrl)), name = "YarnAM")
+  }
+
+  private def checkNumExecutorsFailed() = {
+    if (allocator.getNumExecutorsFailed >= maxNumExecutorFailures) {
+      finish(FinalApplicationStatus.FAILED, "Max number of executor failures reached.")
+
+      val sc = sparkContextRef.get()
+      if (sc != null) {
+        logInfo("Invoking sc stop from checkNumExecutorsFailed")
+        sc.stop()
+      }
+    }
+  }
+
+  /** Add the Yarn IP filter that is required for properly securing the UI. */
+  private def addAmIpFilter() = {
+    val amFilter = "org.apache.hadoop.yarn.server.webproxy.amfilter.AmIpFilter"
+    val proxy = client.getProxyHostAndPort(yarnConf)
+    val parts = proxy.split(":")
+    val proxyBase = System.getenv(ApplicationConstants.APPLICATION_WEB_PROXY_BASE_ENV)
+    val uriBase = "http://" + proxy + proxyBase
+    val params = "PROXY_HOST=" + parts(0) + "," + "PROXY_URI_BASE=" + uriBase
+
+    if (isDriver) {
+      System.setProperty("spark.ui.filters", amFilter)
+      System.setProperty(s"spark.$amFilter.params", params)
+    } else {
+      actor ! AddWebUIFilter(amFilter, params, proxyBase)
+    }
+  }
+
+  private def startUserClass(): Thread = {
+    logInfo("Starting the user JAR in a separate Thread")
+    System.setProperty("spark.executor.instances", args.numExecutors.toString)
+    val mainMethod = Class.forName(args.userClass, false,
+      Thread.currentThread.getContextClassLoader).getMethod("main", classOf[Array[String]])
+
+    val t = new Thread {
+      override def run() {
+        var status = FinalApplicationStatus.FAILED
+        try {
+          // Copy
+          val mainArgs = new Array[String](args.userArgs.size)
+          args.userArgs.copyToArray(mainArgs, 0, args.userArgs.size)
+          mainMethod.invoke(null, mainArgs)
+          // Some apps have "System.exit(0)" at the end.  The user thread will stop here unless
+          // it has an uncaught exception thrown out.  It needs a shutdown hook to set SUCCEEDED.
+          status = FinalApplicationStatus.SUCCEEDED
+        } finally {
+          logDebug("Finishing main")
+        }
+        finalStatus = status
+      }
+    }
+    t.setName("Driver")
+    t.start()
+    t
+  }
+
+  // Actor used to monitor the driver when running in client deploy mode.
+  private class MonitorActor(driverUrl: String) extends Actor {
+
+    var driver: ActorSelection = _
+
+    override def preStart() = {
+      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"
+      context.system.eventStream.subscribe(self, classOf[RemotingLifecycleEvent])
+    }
+
+    override def receive = {
+      case x: DisassociatedEvent =>
+        logInfo(s"Driver terminated or disconnected! Shutting down. $x")
+        finish(FinalApplicationStatus.SUCCEEDED)
+      case x: AddWebUIFilter =>
+        logInfo(s"Add WebUI Filter. $x")
+        driver ! x
+    }
+
+  }
+
+}
+
+object ApplicationMaster extends Logging {
+
+  private var master: ApplicationMaster = _
+
+  def main(args: Array[String]) = {
+    SignalLogger.register(log)
+    val amArgs = new ApplicationMasterArguments(args)
+    SparkHadoopUtil.get.runAsSparkUser { () =>
+      master = new ApplicationMaster(amArgs, new YarnRMClientImpl(amArgs))
+      System.exit(master.run())
+    }
+  }
+
+  private[spark] def sparkContextInitialized(sc: SparkContext) = {
+    master.sparkContextInitialized(sc)
+  }
+
+  private[spark] def sparkContextStopped(sc: SparkContext) = {
+    master.sparkContextStopped(sc)
+  }
+
+}
+
+/**
+ * This object does not provide any special functionality. It exists so that it's easy to tell
+ * apart the client-mode AM from the cluster-mode AM when using tools such as ps or jps.
+ */
+object ExecutorLauncher {
+
+  def main(args: Array[String]) = {
+    ApplicationMaster.main(args)
+  }
+
+}
diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMasterArguments.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMasterArguments.scala
index 424b0fb093..3e6b96fb63 100644
--- a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMasterArguments.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ApplicationMasterArguments.scala
@@ -63,11 +63,6 @@ class ApplicationMasterArguments(val args: Array[String]) {
           executorCores = value
           args = tail
 
-        case Nil =>
-          if (userJar == null || userClass == null) {
-            printUsageAndExit(1)
-          }
-
         case _ =>
           printUsageAndExit(1, args)
       }
@@ -80,16 +75,17 @@ class ApplicationMasterArguments(val args: Array[String]) {
     if (unknownParam != null) {
       System.err.println("Unknown/unsupported param " + unknownParam)
     }
-    System.err.println(
-      "Usage: org.apache.spark.deploy.yarn.ApplicationMaster [options] \n" +
-      "Options:\n" +
-      "  --jar JAR_PATH       Path to your application's JAR file (required)\n" +
-      "  --class CLASS_NAME   Name of your application's main class (required)\n" +
-      "  --args ARGS          Arguments to be passed to your application's main class.\n" +
-      "                       Mutliple invocations are possible, each will be passed in order.\n" +
-      "  --num-executors NUM    Number of executors to start (Default: 2)\n" +
-      "  --executor-cores NUM   Number of cores for the executors (Default: 1)\n" +
-      "  --executor-memory MEM  Memory per executor (e.g. 1000M, 2G) (Default: 1G)\n")
+    System.err.println("""
+      |Usage: org.apache.spark.deploy.yarn.ApplicationMaster [options]
+      |Options:
+      |  --jar JAR_PATH       Path to your application's JAR file
+      |  --class CLASS_NAME   Name of your application's main class
+      |  --args ARGS          Arguments to be passed to your application's main class.
+      |                       Mutliple invocations are possible, each will be passed in order.
+      |  --num-executors NUM    Number of executors to start (Default: 2)
+      |  --executor-cores NUM   Number of cores for the executors (Default: 1)
+      |  --executor-memory MEM  Memory per executor (e.g. 1000M, 2G) (Default: 1G)
+      """.stripMargin)
     System.exit(exitCode)
   }
 }
diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientArguments.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientArguments.scala
index afa4fd4c69..40d8d6d6e6 100644
--- a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientArguments.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientArguments.scala
@@ -37,7 +37,6 @@ class ClientArguments(val args: Array[String], val sparkConf: SparkConf) {
   var numExecutors = 2
   var amQueue = sparkConf.get("QUEUE", "default")
   var amMemory: Int = 512 // MB
-  var amClass: String = "org.apache.spark.deploy.yarn.ApplicationMaster"
   var appName: String = "Spark"
   var priority = 0
 
@@ -78,10 +77,7 @@ class ClientArguments(val args: Array[String], val sparkConf: SparkConf) {
           args = tail
 
         case ("--master-class" | "--am-class") :: value :: tail =>
-          if (args(0) == "--master-class") {
-            println("--master-class is deprecated. Use --am-class instead.")
-          }
-          amClass = value
+          println(s"${args(0)} is deprecated and is not used anymore.")
           args = tail
 
         case ("--master-memory" | "--driver-memory") :: MemoryParam(value) :: tail =>
@@ -133,9 +129,6 @@ class ClientArguments(val args: Array[String], val sparkConf: SparkConf) {
           args = tail
 
         case Nil =>
-          if (userClass == null) {
-            throw new IllegalArgumentException(getUsageMessage())
-          }
 
         case _ =>
           throw new IllegalArgumentException(getUsageMessage(args))
diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala
index 3897b3a373..6cf300c398 100644
--- a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/ClientBase.scala
@@ -42,12 +42,6 @@ import org.apache.spark.{Logging, SecurityManager, SparkConf, SparkContext, Spar
 /**
  * The entry point (starting in Client#main() and Client#run()) for launching Spark on YARN. The
  * Client submits an application to the YARN ResourceManager.
- *
- * Depending on the deployment mode this will launch one of two application master classes:
- * 1. In cluster mode, it will launch an [[org.apache.spark.deploy.yarn.ApplicationMaster]]
- *      which launches a driver program inside of the cluster.
- * 2. In client mode, it will launch an [[org.apache.spark.deploy.yarn.ExecutorLauncher]] to
- *      request executors on behalf of a driver running outside of the cluster.
  */
 trait ClientBase extends Logging {
   val args: ClientArguments
@@ -67,14 +61,11 @@ trait ClientBase extends Logging {
 
   // Additional memory overhead - in mb.
   protected def memoryOverhead: Int = sparkConf.getInt("spark.yarn.driver.memoryOverhead",
-    YarnAllocationHandler.MEMORY_OVERHEAD)
+    YarnSparkHadoopUtil.DEFAULT_MEMORY_OVERHEAD)
 
   // TODO(harvey): This could just go in ClientArguments.
   def validateArgs() = {
     Map(
-      ((args.userJar == null && args.amClass == classOf[ApplicationMaster].getName) ->
-          "Error: You must specify a user jar when running in standalone mode!"),
-      (args.userClass == null) -> "Error: You must specify a user class!",
       (args.numExecutors <= 0) -> "Error: You must specify at least 1 executor!",
       (args.amMemory <= memoryOverhead) -> ("Error: AM memory size must be" +
         "greater than: " + memoryOverhead),
@@ -321,6 +312,8 @@ trait ClientBase extends Logging {
     val amContainer = Records.newRecord(classOf[ContainerLaunchContext])
     amContainer.setLocalResources(localResources)
 
+    val isLaunchingDriver = args.userClass != null
+
     // In cluster mode, if the deprecated SPARK_JAVA_OPTS is set, we need to propagate it to
     // executors. But we can't just set spark.executor.extraJavaOptions, because the driver's
     // SparkContext will not let that set spark* system properties, which is expected behavior for
@@ -329,7 +322,7 @@ trait ClientBase extends Logging {
     // Note that to warn the user about the deprecation in cluster mode, some code from
     // SparkConf#validateSettings() is duplicated here (to avoid triggering the condition
     // described above).
-    if (args.amClass == classOf[ApplicationMaster].getName) {
+    if (isLaunchingDriver) {
       sys.env.get("SPARK_JAVA_OPTS").foreach { value =>
         val warning =
           s"""
@@ -389,7 +382,7 @@ trait ClientBase extends Logging {
       javaOpts += YarnSparkHadoopUtil.escapeForShell(s"-D$k=$v")
     }
 
-    if (args.amClass == classOf[ApplicationMaster].getName) {
+    if (isLaunchingDriver) {
       sparkConf.getOption("spark.driver.extraJavaOptions")
         .orElse(sys.env.get("SPARK_JAVA_OPTS"))
         .foreach(opts => javaOpts += opts)
@@ -397,22 +390,37 @@ trait ClientBase extends Logging {
         .foreach(p => javaOpts += s"-Djava.library.path=$p")
     }
 
-    // Command for the ApplicationMaster
-    val commands = Seq(Environment.JAVA_HOME.$() + "/bin/java", "-server") ++
-      javaOpts ++
-      Seq(args.amClass, "--class", YarnSparkHadoopUtil.escapeForShell(args.userClass),
-        "--jar ", YarnSparkHadoopUtil.escapeForShell(args.userJar),
-        userArgsToString(args),
-        "--executor-memory", args.executorMemory.toString,
+    val userClass =
+      if (args.userClass != null) {
+        Seq("--class", YarnSparkHadoopUtil.escapeForShell(args.userClass))
+      } else {
+        Nil
+      }
+    val amClass =
+      if (isLaunchingDriver) {
+        classOf[ApplicationMaster].getName()
+      } else {
+        classOf[ApplicationMaster].getName().replace("ApplicationMaster", "ExecutorLauncher")
+      }
+    val amArgs =
+      Seq(amClass) ++ userClass ++
+      (if (args.userJar != null) Seq("--jar", args.userJar) else Nil) ++
+      Seq("--executor-memory", args.executorMemory.toString,
         "--executor-cores", args.executorCores.toString,
         "--num-executors ", args.numExecutors.toString,
+        userArgsToString(args))
+
+    // Command for the ApplicationMaster
+    val commands = Seq(Environment.JAVA_HOME.$() + "/bin/java", "-server") ++
+      javaOpts ++ amArgs ++
+      Seq(
         "1>", ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stdout",
         "2>", ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stderr")
 
     logInfo("Yarn AM launch context:")
-    logInfo(s"  class:   ${args.amClass}")
-    logInfo(s"  env:     $env")
-    logInfo(s"  command: ${commands.mkString(" ")}")
+    logInfo(s"  user class: ${args.userClass}")
+    logInfo(s"  env:        $env")
+    logInfo(s"  command:    ${commands.mkString(" ")}")
 
     // TODO: it would be nicer to just make sure there are no null commands here
     val printableCommands = commands.map(s => if (s == null) "null" else s).toList
@@ -623,7 +631,7 @@ object ClientBase extends Logging {
     YarnSparkHadoopUtil.addToEnvironment(env, Environment.CLASSPATH.name, path,
             File.pathSeparator)
 
-  /** 
+  /**
    * Get the list of namenodes the user may access.
    */
   private[yarn] def getNameNodesToAccess(sparkConf: SparkConf): Set[Path] = {
diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala
new file mode 100644
index 0000000000..cad94e5e19
--- /dev/null
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocator.scala
@@ -0,0 +1,34 @@
+/*
+ * 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.deploy.yarn
+
+object AllocationType extends Enumeration {
+  type AllocationType = Value
+  val HOST, RACK, ANY = Value
+}
+
+/**
+ * Interface that defines a Yarn allocator.
+ */
+trait YarnAllocator {
+
+  def allocateResources(): Unit
+  def getNumExecutorsFailed: Int
+  def getNumExecutorsRunning: Int
+
+}
diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnRMClient.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnRMClient.scala
new file mode 100644
index 0000000000..922d7d1a85
--- /dev/null
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnRMClient.scala
@@ -0,0 +1,67 @@
+/*
+ * 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.deploy.yarn
+
+import scala.collection.{Map, Set}
+
+import org.apache.hadoop.yarn.conf.YarnConfiguration
+import org.apache.hadoop.yarn.api.records._
+
+import org.apache.spark.{SparkConf, SparkContext}
+import org.apache.spark.scheduler.SplitInfo
+
+/**
+ * Interface that defines a Yarn RM client. Abstracts away Yarn version-specific functionality that
+ * is used by Spark's AM.
+ */
+trait YarnRMClient {
+
+  /**
+   * Registers the application master with the RM.
+   *
+   * @param conf The Yarn configuration.
+   * @param sparkConf The Spark configuration.
+   * @param preferredNodeLocations Map with hints about where to allocate containers.
+   * @param uiAddress Address of the SparkUI.
+   * @param uiHistoryAddress Address of the application on the History Server.
+   */
+  def register(
+      conf: YarnConfiguration,
+      sparkConf: SparkConf,
+      preferredNodeLocations: Map[String, Set[SplitInfo]],
+      uiAddress: String,
+      uiHistoryAddress: String): YarnAllocator
+
+  /**
+   * Shuts down the AM. Guaranteed to only be called once.
+   *
+   * @param status The final status of the AM.
+   * @param diagnostics Diagnostics message to include in the final status.
+   */
+  def shutdown(status: FinalApplicationStatus, diagnostics: String = ""): Unit
+
+  /** Returns the attempt ID. */
+  def getAttemptId(): ApplicationAttemptId
+
+  /** Returns the RM's proxy host and port. */
+  def getProxyHostAndPort(conf: YarnConfiguration): String
+
+  /** Returns the maximum number of attempts to register the AM. */
+  def getMaxRegAttempts(conf: YarnConfiguration): Int
+
+}
diff --git a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtil.scala b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtil.scala
index 10aef5eb24..2aa27a1908 100644
--- a/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtil.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/deploy/yarn/YarnSparkHadoopUtil.scala
@@ -17,8 +17,11 @@
 
 package org.apache.spark.deploy.yarn
 
+import java.lang.{Boolean => JBoolean}
+import java.util.{Collections, Set => JSet}
 import java.util.regex.Matcher
 import java.util.regex.Pattern
+import java.util.concurrent.ConcurrentHashMap
 
 import scala.collection.mutable.HashMap
 
@@ -29,11 +32,13 @@ import org.apache.hadoop.security.UserGroupInformation
 import org.apache.hadoop.util.StringInterner
 import org.apache.hadoop.yarn.conf.YarnConfiguration
 import org.apache.hadoop.yarn.api.ApplicationConstants
+import org.apache.hadoop.yarn.util.RackResolver
 import org.apache.hadoop.conf.Configuration
 
 import org.apache.spark.{SparkConf, SparkContext}
 import org.apache.spark.deploy.history.HistoryServer
 import org.apache.spark.deploy.SparkHadoopUtil
+import org.apache.spark.util.Utils
 
 /**
  * Contains util methods to interact with Hadoop from spark.
@@ -79,6 +84,21 @@ class YarnSparkHadoopUtil extends SparkHadoopUtil {
 }
 
 object YarnSparkHadoopUtil {
+  // Additional memory overhead - in mb.
+  val DEFAULT_MEMORY_OVERHEAD = 384
+
+  val ANY_HOST = "*"
+
+  // All RM requests are issued with same priority : we do not (yet) have any distinction between
+  // request types (like map/reduce in hadoop for example)
+  val RM_REQUEST_PRIORITY = 1
+
+  // Host to rack map - saved from allocation requests. We are expecting this not to change.
+  // Note that it is possible for this to change : and ResourceManager will indicate that to us via
+  // update response to allocate. But we are punting on handling that for now.
+  private val hostToRack = new ConcurrentHashMap[String, String]()
+  private val rackToHostSet = new ConcurrentHashMap[String, JSet[String]]()
+
   def addToEnvironment(
       env: HashMap[String, String],
       variable: String,
@@ -173,4 +193,35 @@ object YarnSparkHadoopUtil {
     }
   }
 
+  private[spark] def lookupRack(conf: Configuration, host: String): String = {
+    if (!hostToRack.contains(host)) {
+      populateRackInfo(conf, host)
+    }
+    hostToRack.get(host)
+  }
+
+  private[spark] def populateRackInfo(conf: Configuration, hostname: String) {
+    Utils.checkHost(hostname)
+
+    if (!hostToRack.containsKey(hostname)) {
+      // If there are repeated failures to resolve, all to an ignore list.
+      val rackInfo = RackResolver.resolve(conf, hostname)
+      if (rackInfo != null && rackInfo.getNetworkLocation != null) {
+        val rack = rackInfo.getNetworkLocation
+        hostToRack.put(hostname, rack)
+        if (! rackToHostSet.containsKey(rack)) {
+          rackToHostSet.putIfAbsent(rack,
+            Collections.newSetFromMap(new ConcurrentHashMap[String, JBoolean]()))
+        }
+        rackToHostSet.get(rack).add(hostname)
+
+        // TODO(harvey): Figure out what this comment means...
+        // Since RackResolver caches, we are disabling this for now ...
+      } /* else {
+        // right ? Else we will keep calling rack resolver in case we cant resolve rack info ...
+        hostToRack.put(hostname, null)
+      } */
+    }
+  }
+
 }
diff --git a/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientClusterScheduler.scala b/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientClusterScheduler.scala
index 3474112ded..d162b4c433 100644
--- a/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientClusterScheduler.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientClusterScheduler.scala
@@ -19,22 +19,21 @@ package org.apache.spark.scheduler.cluster
 
 import org.apache.spark._
 import org.apache.hadoop.conf.Configuration
-import org.apache.spark.deploy.yarn.YarnAllocationHandler
+import org.apache.spark.deploy.yarn.YarnSparkHadoopUtil
 import org.apache.spark.scheduler.TaskSchedulerImpl
 import org.apache.spark.util.Utils
 
 /**
- *
- * This scheduler launches executors through Yarn - by calling into Client to launch ExecutorLauncher as AM.
+ * This scheduler launches executors through Yarn - by calling into Client to launch the Spark AM.
  */
-private[spark] class YarnClientClusterScheduler(sc: SparkContext, conf: Configuration) extends TaskSchedulerImpl(sc) {
+private[spark] class YarnClientClusterScheduler(sc: SparkContext, conf: Configuration)
+  extends TaskSchedulerImpl(sc) {
 
   def this(sc: SparkContext) = this(sc, new Configuration())
 
   // By default, rack is unknown
   override def getRackForHost(hostPort: String): Option[String] = {
     val host = Utils.parseHostPort(hostPort)._1
-    val retval = YarnAllocationHandler.lookupRack(conf, host)
-    if (retval != null) Some(retval) else None
+    Option(YarnSparkHadoopUtil.lookupRack(conf, host))
   }
 }
diff --git a/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientSchedulerBackend.scala b/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientSchedulerBackend.scala
index 833e249f9f..a5f537dd9d 100644
--- a/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientSchedulerBackend.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientSchedulerBackend.scala
@@ -19,7 +19,7 @@ package org.apache.spark.scheduler.cluster
 
 import org.apache.hadoop.yarn.api.records.{ApplicationId, YarnApplicationState}
 import org.apache.spark.{SparkException, Logging, SparkContext}
-import org.apache.spark.deploy.yarn.{Client, ClientArguments, ExecutorLauncher, YarnSparkHadoopUtil}
+import org.apache.spark.deploy.yarn.{Client, ClientArguments, YarnSparkHadoopUtil}
 import org.apache.spark.scheduler.TaskSchedulerImpl
 
 import scala.collection.mutable.ArrayBuffer
@@ -60,10 +60,7 @@ private[spark] class YarnClientSchedulerBackend(
 
     val argsArrayBuf = new ArrayBuffer[String]()
     argsArrayBuf += (
-      "--class", "notused",
-      "--jar", null, // The primary jar will be added dynamically in SparkContext.
-      "--args", hostport,
-      "--am-class", classOf[ExecutorLauncher].getName
+      "--args", hostport
     )
 
     // process any optional arguments, given either as environment variables
diff --git a/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClusterScheduler.scala b/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClusterScheduler.scala
index 9aeca4a637..69f40225a2 100644
--- a/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClusterScheduler.scala
+++ b/yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClusterScheduler.scala
@@ -18,16 +18,17 @@
 package org.apache.spark.scheduler.cluster
 
 import org.apache.spark._
-import org.apache.spark.deploy.yarn.{ApplicationMaster, YarnAllocationHandler}
+import org.apache.spark.deploy.yarn.{ApplicationMaster, YarnSparkHadoopUtil}
 import org.apache.spark.scheduler.TaskSchedulerImpl
 import org.apache.spark.util.Utils
 import org.apache.hadoop.conf.Configuration
 
 /**
- *
- * This is a simple extension to ClusterScheduler - to ensure that appropriate initialization of ApplicationMaster, etc is done
+ * This is a simple extension to ClusterScheduler - to ensure that appropriate initialization of
+ * ApplicationMaster, etc is done
  */
-private[spark] class YarnClusterScheduler(sc: SparkContext, conf: Configuration) extends TaskSchedulerImpl(sc) {
+private[spark] class YarnClusterScheduler(sc: SparkContext, conf: Configuration)
+  extends TaskSchedulerImpl(sc) {
 
   logInfo("Created YarnClusterScheduler")
 
@@ -42,7 +43,7 @@ private[spark] class YarnClusterScheduler(sc: SparkContext, conf: Configuration)
   // By default, rack is unknown
   override def getRackForHost(hostPort: String): Option[String] = {
     val host = Utils.parseHostPort(hostPort)._1
-    val retval = YarnAllocationHandler.lookupRack(conf, host)
+    val retval = YarnSparkHadoopUtil.lookupRack(conf, host)
     if (retval != null) Some(retval) else None
   }
 
@@ -51,4 +52,10 @@ private[spark] class YarnClusterScheduler(sc: SparkContext, conf: Configuration)
     super.postStartHook()
     logInfo("YarnClusterScheduler.postStartHook done")
   }
+
+  override def stop() {
+    super.stop()
+    ApplicationMaster.sparkContextStopped(sc)
+  }
+
 }