rebase to master

Author: sarutak

.gitignore

@@ -51,6 +51,7 @@ unit-tests.log
 rat-results.txt
 scalastyle.txt
 conf/*.conf
+scalastyle-output.xml
 
 # For Hive
 metastore_db/

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/Aggregator.scala

@@ -55,10 +55,7 @@ case class Aggregator[K, V, C] (
       combiners.iterator
     } else {
       val combiners = new ExternalAppendOnlyMap[K, V, C](createCombiner, mergeValue, mergeCombiners)
-      while (iter.hasNext) {
-        val pair = iter.next()
-        combiners.insert(pair._1, pair._2)
-      }
+      combiners.insertAll(iter)
       // TODO: Make this non optional in a future release
       Option(context).foreach(c => c.taskMetrics.memoryBytesSpilled = combiners.memoryBytesSpilled)
       Option(context).foreach(c => c.taskMetrics.diskBytesSpilled = combiners.diskBytesSpilled)

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

@@ -17,9 +17,9 @@
 
 package org.apache.spark
 
-import scala.collection.mutable.{ArrayBuffer, HashSet}
+import scala.collection.mutable
+import scala.collection.mutable.ArrayBuffer
 
-import org.apache.spark.executor.InputMetrics
 import org.apache.spark.rdd.RDD
 import org.apache.spark.storage._
 
@@ -30,7 +30,7 @@ import org.apache.spark.storage._
 private[spark] class CacheManager(blockManager: BlockManager) extends Logging {
 
   /** Keys of RDD partitions that are being computed/loaded. */
-  private val loading = new HashSet[RDDBlockId]()
+  private val loading = new mutable.HashSet[RDDBlockId]
 
   /** Gets or computes an RDD partition. Used by RDD.iterator() when an RDD is cached. */
   def getOrCompute[T](
@@ -118,36 +118,66 @@ private[spark] class CacheManager(blockManager: BlockManager) extends Logging {
   }
 
   /**
-   * Cache the values of a partition, keeping track of any updates in the storage statuses
-   * of other blocks along the way.
+   * Cache the values of a partition, keeping track of any updates in the storage statuses of
+   * other blocks along the way.
+   *
+   * The effective storage level refers to the level that actually specifies BlockManager put
+   * behavior, not the level originally specified by the user. This is mainly for forcing a
+   * MEMORY_AND_DISK partition to disk if there is not enough room to unroll the partition,
+   * while preserving the the original semantics of the RDD as specified by the application.
    */
   private def putInBlockManager[T](
       key: BlockId,
       values: Iterator[T],
-      storageLevel: StorageLevel,
-      updatedBlocks: ArrayBuffer[(BlockId, BlockStatus)]): Iterator[T] = {
-
-    if (!storageLevel.useMemory) {
-      /* This RDD is not to be cached in memory, so we can just pass the computed values
-       * as an iterator directly to the BlockManager, rather than first fully unrolling
-       * it in memory. The latter option potentially uses much more memory and risks OOM
-       * exceptions that can be avoided. */
-      updatedBlocks ++= blockManager.put(key, values, storageLevel, tellMaster = true)
+      level: StorageLevel,
+      updatedBlocks: ArrayBuffer[(BlockId, BlockStatus)],
+      effectiveStorageLevel: Option[StorageLevel] = None): Iterator[T] = {
+
+    val putLevel = effectiveStorageLevel.getOrElse(level)
+    if (!putLevel.useMemory) {
+      /*
+       * This RDD is not to be cached in memory, so we can just pass the computed values as an
+       * iterator directly to the BlockManager rather than first fully unrolling it in memory.
+       */
+      updatedBlocks ++=
+        blockManager.putIterator(key, values, level, tellMaster = true, effectiveStorageLevel)
       blockManager.get(key) match {
         case Some(v) => v.data.asInstanceOf[Iterator[T]]
         case None =>
           logInfo(s"Failure to store $key")
           throw new BlockException(key, s"Block manager failed to return cached value for $key!")
       }
     } else {
-      /* This RDD is to be cached in memory. In this case we cannot pass the computed values
+      /*
+       * This RDD is to be cached in memory. In this case we cannot pass the computed values
        * to the BlockManager as an iterator and expect to read it back later. This is because
-       * we may end up dropping a partition from memory store before getting it back, e.g.
-       * when the entirety of the RDD does not fit in memory. */
-      val elements = new ArrayBuffer[Any]
-      elements ++= values
-      updatedBlocks ++= blockManager.put(key, elements, storageLevel, tellMaster = true)
-      elements.iterator.asInstanceOf[Iterator[T]]
+       * we may end up dropping a partition from memory store before getting it back.
+       *
+       * In addition, we must be careful to not unroll the entire partition in memory at once.
+       * Otherwise, we may cause an OOM exception if the JVM does not have enough space for this
+       * single partition. Instead, we unroll the values cautiously, potentially aborting and
+       * dropping the partition to disk if applicable.
+       */
+      blockManager.memoryStore.unrollSafely(key, values, updatedBlocks) match {
+        case Left(arr) =>
+          // We have successfully unrolled the entire partition, so cache it in memory
+          updatedBlocks ++=
+            blockManager.putArray(key, arr, level, tellMaster = true, effectiveStorageLevel)
+          arr.iterator.asInstanceOf[Iterator[T]]
+        case Right(it) =>
+          // There is not enough space to cache this partition in memory
+          logWarning(s"Not enough space to cache partition $key in memory! " +
+            s"Free memory is ${blockManager.memoryStore.freeMemory} bytes.")
+          val returnValues = it.asInstanceOf[Iterator[T]]
+          if (putLevel.useDisk) {
+            logWarning(s"Persisting partition $key to disk instead.")
+            val diskOnlyLevel = StorageLevel(useDisk = true, useMemory = false,
+              useOffHeap = false, deserialized = false, putLevel.replication)
+            putInBlockManager[T](key, returnValues, level, updatedBlocks, Some(diskOnlyLevel))
+          } else {
+            returnValues
+          }
+      }
     }
   }
 

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

@@ -19,6 +19,7 @@ package org.apache.spark
 
 import org.apache.spark.annotation.DeveloperApi
 import org.apache.spark.rdd.RDD
+import org.apache.spark.rdd.SortOrder.SortOrder
 import org.apache.spark.serializer.Serializer
 import org.apache.spark.shuffle.ShuffleHandle
 
@@ -62,7 +63,8 @@ class ShuffleDependency[K, V, C](
     val serializer: Option[Serializer] = None,
     val keyOrdering: Option[Ordering[K]] = None,
     val aggregator: Option[Aggregator[K, V, C]] = None,
-    val mapSideCombine: Boolean = false)
+    val mapSideCombine: Boolean = false,
+    val sortOrder: Option[SortOrder] = None)
   extends Dependency(rdd.asInstanceOf[RDD[Product2[K, V]]]) {
 
   val shuffleId: Int = rdd.context.newShuffleId()

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

@@ -1037,7 +1037,7 @@ class SparkContext(config: SparkConf) extends Logging {
    */
   private[spark] def getCallSite(): CallSite = {
     Option(getLocalProperty("externalCallSite")) match {
-      case Some(callSite) => CallSite(callSite, long = "")
+      case Some(callSite) => CallSite(callSite, longForm = "")
       case None => Utils.getCallSite
     }
   }
@@ -1059,11 +1059,12 @@ class SparkContext(config: SparkConf) extends Logging {
     }
     val callSite = getCallSite
     val cleanedFunc = clean(func)
-    logInfo("Starting job: " + callSite.short)
+    logInfo("Starting job: " + callSite.shortForm)
     val start = System.nanoTime
     dagScheduler.runJob(rdd, cleanedFunc, partitions, callSite, allowLocal,
       resultHandler, localProperties.get)
-    logInfo("Job finished: " + callSite.short + ", took " + (System.nanoTime - start) / 1e9 + " s")
+    logInfo(
+      "Job finished: " + callSite.shortForm + ", took " + (System.nanoTime - start) / 1e9 + " s")
     rdd.doCheckpoint()
   }
 
@@ -1144,11 +1145,12 @@ class SparkContext(config: SparkConf) extends Logging {
       evaluator: ApproximateEvaluator[U, R],
       timeout: Long): PartialResult[R] = {
     val callSite = getCallSite
-    logInfo("Starting job: " + callSite.short)
+    logInfo("Starting job: " + callSite.shortForm)
     val start = System.nanoTime
     val result = dagScheduler.runApproximateJob(rdd, func, evaluator, callSite, timeout,
       localProperties.get)
-    logInfo("Job finished: " + callSite.short + ", took " + (System.nanoTime - start) / 1e9 + " s")
+    logInfo(
+      "Job finished: " + callSite.shortForm + ", took " + (System.nanoTime - start) / 1e9 + " s")
     result
   }
 

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

@@ -67,7 +67,7 @@ class SparkEnv (
     val metricsSystem: MetricsSystem,
     val conf: SparkConf) extends Logging {
 
-  // A mapping of thread ID to amount of memory used for shuffle in bytes
+  // A mapping of thread ID to amount of memory, in bytes, used for shuffle aggregations
   // All accesses should be manually synchronized
   val shuffleMemoryMap = mutable.HashMap[Long, Long]()
 

core/src/main/scala/org/apache/spark/api/java/JavaSparkContext.scala

@@ -34,7 +34,7 @@ import org.apache.spark._
 import org.apache.spark.SparkContext.{DoubleAccumulatorParam, IntAccumulatorParam}
 import org.apache.spark.api.java.JavaSparkContext.fakeClassTag
 import org.apache.spark.broadcast.Broadcast
-import org.apache.spark.rdd.RDD
+import org.apache.spark.rdd.{EmptyRDD, RDD}
 
 /**
  * A Java-friendly version of [[org.apache.spark.SparkContext]] that returns
@@ -112,6 +112,9 @@ class JavaSparkContext(val sc: SparkContext) extends JavaSparkContextVarargsWork
 
   def startTime: java.lang.Long = sc.startTime
 
+  /** The version of Spark on which this application is running. */
+  def version: String = sc.version
+
   /** Default level of parallelism to use when not given by user (e.g. parallelize and makeRDD). */
   def defaultParallelism: java.lang.Integer = sc.defaultParallelism
 
@@ -132,6 +135,13 @@ class JavaSparkContext(val sc: SparkContext) extends JavaSparkContextVarargsWork
     sc.parallelize(JavaConversions.asScalaBuffer(list), numSlices)
   }
 
+  /** Get an RDD that has no partitions or elements. */
+  def emptyRDD[T]: JavaRDD[T] = {
+    implicit val ctag: ClassTag[T] = fakeClassTag
+    JavaRDD.fromRDD(new EmptyRDD[T](sc))
+  }
+
+
   /** Distribute a local Scala collection to form an RDD. */
   def parallelize[T](list: java.util.List[T]): JavaRDD[T] =
     parallelize(list, sc.defaultParallelism)

core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala

@@ -37,8 +37,8 @@ import org.apache.spark.broadcast.Broadcast
 import org.apache.spark.rdd.RDD
 import org.apache.spark.util.Utils
 
-private[spark] class PythonRDD[T: ClassTag](
-    parent: RDD[T],
+private[spark] class PythonRDD(
+    parent: RDD[_],
     command: Array[Byte],
     envVars: JMap[String, String],
     pythonIncludes: JList[String],
@@ -57,7 +57,10 @@ private[spark] class PythonRDD[T: ClassTag](
   override def compute(split: Partition, context: TaskContext): Iterator[Array[Byte]] = {
     val startTime = System.currentTimeMillis
     val env = SparkEnv.get
-    val worker: Socket = env.createPythonWorker(pythonExec, envVars.toMap)
+    val localdir = env.blockManager.diskBlockManager.localDirs.map(
+      f => f.getPath()).mkString(",")
+    val worker: Socket = env.createPythonWorker(pythonExec,
+      envVars.toMap + ("SPARK_LOCAL_DIR" -> localdir))
 
     // Start a thread to feed the process input from our parent's iterator
     val writerThread = new WriterThread(env, worker, split, context)

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()