Merge remote-tracking branch 'upstream/master' into dataTypeAndSchema

Conflicts:
	sql/hive/src/main/scala/org/apache/spark/sql/hive/hiveUdfs.scala

Author: yhuai

assembly/pom.xml

@@ -39,6 +39,7 @@
     <deb.pkg.name>spark</deb.pkg.name>
     <deb.install.path>/usr/share/spark</deb.install.path>
     <deb.user>root</deb.user>
+    <deb.bin.filemode>744</deb.bin.filemode>
   </properties>
 
   <dependencies>
@@ -276,7 +277,7 @@
                         <user>${deb.user}</user>
                         <group>${deb.user}</group>
                         <prefix>${deb.install.path}/bin</prefix>
-                        <filemode>744</filemode>
+                        <filemode>${deb.bin.filemode}</filemode>
                       </mapper>
                     </data>
                     <data>

bagel/src/main/scala/org/apache/spark/bagel/Bagel.scala

@@ -72,6 +72,7 @@ object Bagel extends Logging {
     var verts = vertices
     var msgs = messages
     var noActivity = false
+    var lastRDD: RDD[(K, (V, Array[M]))] = null
     do {
       logInfo("Starting superstep " + superstep + ".")
       val startTime = System.currentTimeMillis
@@ -83,6 +84,10 @@ object Bagel extends Logging {
       val superstep_ = superstep  // Create a read-only copy of superstep for capture in closure
       val (processed, numMsgs, numActiveVerts) =
         comp[K, V, M, C](sc, grouped, compute(_, _, aggregated, superstep_), storageLevel)
+      if (lastRDD != null) {
+        lastRDD.unpersist(false)
+      }
+      lastRDD = processed
 
       val timeTaken = System.currentTimeMillis - startTime
       logInfo("Superstep %d took %d s".format(superstep, timeTaken / 1000))

core/src/main/scala/org/apache/spark/deploy/SparkSubmit.scala

@@ -269,6 +269,9 @@ object SparkSubmit {
       sysProps.getOrElseUpdate(k, v)
     }
 
+    // Spark properties included on command line take precedence
+    sysProps ++= args.sparkProperties
+
     (childArgs, childClasspath, sysProps, childMainClass)
   }
 

core/src/main/scala/org/apache/spark/deploy/SparkSubmitArguments.scala

@@ -55,6 +55,7 @@ private[spark] class SparkSubmitArguments(args: Seq[String]) {
   var verbose: Boolean = false
   var isPython: Boolean = false
   var pyFiles: String = null
+  val sparkProperties: HashMap[String, String] = new HashMap[String, String]()
 
   parseOpts(args.toList)
   loadDefaults()
@@ -177,6 +178,7 @@ private[spark] class SparkSubmitArguments(args: Seq[String]) {
     |  executorCores           $executorCores
     |  totalExecutorCores      $totalExecutorCores
     |  propertiesFile          $propertiesFile
+    |  extraSparkProperties    $sparkProperties
     |  driverMemory            $driverMemory
     |  driverCores             $driverCores
     |  driverExtraClassPath    $driverExtraClassPath
@@ -290,6 +292,13 @@ private[spark] class SparkSubmitArguments(args: Seq[String]) {
         jars = Utils.resolveURIs(value)
         parse(tail)
 
+      case ("--conf" | "-c") :: value :: tail =>
+        value.split("=", 2).toSeq match {
+          case Seq(k, v) => sparkProperties(k) = v
+          case _ => SparkSubmit.printErrorAndExit(s"Spark config without '=': $value")
+        }
+        parse(tail)
+
       case ("--help" | "-h") :: tail =>
         printUsageAndExit(0)
 
@@ -349,6 +358,8 @@ private[spark] class SparkSubmitArguments(args: Seq[String]) {
         |                              on the PYTHONPATH for Python apps.
         |  --files FILES               Comma-separated list of files to be placed in the working
         |                              directory of each executor.
+        |
+        |  --conf PROP=VALUE           Arbitrary Spark configuration property.
         |  --properties-file FILE      Path to a file from which to load extra properties. If not
         |                              specified, this will look for conf/spark-defaults.conf.
         |

core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala

@@ -169,7 +169,8 @@ private[history] class FsHistoryProvider(conf: SparkConf) extends ApplicationHis
     val ui: SparkUI = if (renderUI) {
         val conf = this.conf.clone()
         val appSecManager = new SecurityManager(conf)
-        new SparkUI(conf, appSecManager, replayBus, appId, "/history/" + appId)
+        new SparkUI(conf, appSecManager, replayBus, appId,
+          HistoryServer.UI_PATH_PREFIX + s"/$appId")
         // Do not call ui.bind() to avoid creating a new server for each application
       } else {
         null

core/src/main/scala/org/apache/spark/deploy/history/HistoryPage.scala

@@ -75,7 +75,7 @@ private[spark] class HistoryPage(parent: HistoryServer) extends WebUIPage("") {
     "Last Updated")
 
   private def appRow(info: ApplicationHistoryInfo): Seq[Node] = {
-    val uiAddress = "/history/" + info.id
+    val uiAddress = HistoryServer.UI_PATH_PREFIX + s"/${info.id}"
     val startTime = UIUtils.formatDate(info.startTime)
     val endTime = UIUtils.formatDate(info.endTime)
     val duration = UIUtils.formatDuration(info.endTime - info.startTime)

core/src/main/scala/org/apache/spark/deploy/history/HistoryServer.scala

@@ -114,7 +114,7 @@ class HistoryServer(
     attachHandler(createStaticHandler(SparkUI.STATIC_RESOURCE_DIR, "/static"))
 
     val contextHandler = new ServletContextHandler
-    contextHandler.setContextPath("/history")
+    contextHandler.setContextPath(HistoryServer.UI_PATH_PREFIX)
     contextHandler.addServlet(new ServletHolder(loaderServlet), "/*")
     attachHandler(contextHandler)
   }
@@ -172,6 +172,8 @@ class HistoryServer(
 object HistoryServer extends Logging {
   private val conf = new SparkConf
 
+  val UI_PATH_PREFIX = "/history"
+
   def main(argStrings: Array[String]) {
     SignalLogger.register(log)
     initSecurity()

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

@@ -35,6 +35,7 @@ import akka.serialization.SerializationExtension
 import org.apache.spark.{Logging, SecurityManager, SparkConf, SparkException}
 import org.apache.spark.deploy.{ApplicationDescription, DriverDescription, ExecutorState}
 import org.apache.spark.deploy.DeployMessages._
+import org.apache.spark.deploy.history.HistoryServer
 import org.apache.spark.deploy.master.DriverState.DriverState
 import org.apache.spark.deploy.master.MasterMessages._
 import org.apache.spark.deploy.master.ui.MasterWebUI
@@ -664,9 +665,10 @@ private[spark] class Master(
    */
   def rebuildSparkUI(app: ApplicationInfo): Boolean = {
     val appName = app.desc.name
+    val notFoundBasePath = HistoryServer.UI_PATH_PREFIX + "/not-found"
     val eventLogDir = app.desc.eventLogDir.getOrElse {
       // Event logging is not enabled for this application
-      app.desc.appUiUrl = "/history/not-found"
+      app.desc.appUiUrl = notFoundBasePath
       return false
     }
     val fileSystem = Utils.getHadoopFileSystem(eventLogDir)
@@ -681,13 +683,14 @@ private[spark] class Master(
       logWarning(msg)
       msg += " Did you specify the correct logging directory?"
       msg = URLEncoder.encode(msg, "UTF-8")
-      app.desc.appUiUrl = s"/history/not-found?msg=$msg&title=$title"
+      app.desc.appUiUrl = notFoundBasePath + s"?msg=$msg&title=$title"
       return false
     }
 
     try {
       val replayBus = new ReplayListenerBus(eventLogPaths, fileSystem, compressionCodec)
-      val ui = new SparkUI(new SparkConf, replayBus, appName + " (completed)", "/history/" + app.id)
+      val ui = new SparkUI(new SparkConf, replayBus, appName + " (completed)",
+        HistoryServer.UI_PATH_PREFIX + s"/${app.id}")
       replayBus.replay()
       appIdToUI(app.id) = ui
       webUi.attachSparkUI(ui)
@@ -702,7 +705,7 @@ private[spark] class Master(
         var msg = s"Exception in replaying log for application $appName!"
         logError(msg, e)
         msg = URLEncoder.encode(msg, "UTF-8")
-        app.desc.appUiUrl = s"/history/not-found?msg=$msg&exception=$exception&title=$title"
+        app.desc.appUiUrl = notFoundBasePath + s"?msg=$msg&exception=$exception&title=$title"
         false
     }
   }

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

@@ -63,6 +63,13 @@ private[spark] class EventLoggingListener(
   // For testing. Keep track of all JSON serialized events that have been logged.
   private[scheduler] val loggedEvents = new ArrayBuffer[JValue]
 
+  /**
+   * Return only the unique application directory without the base directory.
+   */
+  def getApplicationLogDir(): String = {
+    name
+  }
+
   /**
    * Begin logging events.
    * If compression is used, log a file that indicates which compression library is used.

core/src/main/scala/org/apache/spark/scheduler/local/LocalBackend.scala

@@ -57,7 +57,7 @@ private[spark] class LocalActor(
     case StatusUpdate(taskId, state, serializedData) =>
       scheduler.statusUpdate(taskId, state, serializedData)
       if (TaskState.isFinished(state)) {
-        freeCores += 1
+        freeCores += scheduler.CPUS_PER_TASK
         reviveOffers()
       }
 
@@ -68,7 +68,7 @@ private[spark] class LocalActor(
   def reviveOffers() {
     val offers = Seq(new WorkerOffer(localExecutorId, localExecutorHostname, freeCores))
     for (task <- scheduler.resourceOffers(offers).flatten) {
-      freeCores -= 1
+      freeCores -= scheduler.CPUS_PER_TASK
       executor.launchTask(executorBackend, task.taskId, task.name, task.serializedTask)
     }
   }