Merge branch 'master' into sqlconf

Author: zsxwing

.rat-excludes

@@ -28,6 +28,7 @@ spark-env.sh
 spark-env.cmd
 spark-env.sh.template
 log4j-defaults.properties
+log4j-defaults-repl.properties
 bootstrap-tooltip.js
 jquery-1.11.1.min.js
 d3.min.js

core/src/main/resources/org/apache/spark/log4j-defaults-repl.properties

@@ -0,0 +1,12 @@
+# Set everything to be logged to the console
+log4j.rootCategory=WARN, console
+log4j.appender.console=org.apache.log4j.ConsoleAppender
+log4j.appender.console.target=System.err
+log4j.appender.console.layout=org.apache.log4j.PatternLayout
+log4j.appender.console.layout.ConversionPattern=%d{yy/MM/dd HH:mm:ss} %p %c{1}: %m%n
+
+# Settings to quiet third party logs that are too verbose
+log4j.logger.org.spark-project.jetty=WARN
+log4j.logger.org.spark-project.jetty.util.component.AbstractLifeCycle=ERROR
+log4j.logger.org.apache.spark.repl.SparkIMain$exprTyper=INFO
+log4j.logger.org.apache.spark.repl.SparkILoop$SparkILoopInterpreter=INFO

core/src/main/resources/org/apache/spark/ui/static/spark-dag-viz.js

@@ -235,7 +235,7 @@ function renderDagVizForJob(svgContainer) {
     // them separately later. Note that we cannot draw them now because we need to
     // put these edges in a separate container that is on top of all stage graphs.
     metadata.selectAll(".incoming-edge").each(function(v) {
-      var edge = d3.select(this).text().split(","); // e.g. 3,4 => [3, 4]
+      var edge = d3.select(this).text().trim().split(","); // e.g. 3,4 => [3, 4]
       crossStageEdges.push(edge);
     });
   });

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

@@ -121,13 +121,25 @@ trait Logging {
     if (usingLog4j12) {
       val log4j12Initialized = LogManager.getRootLogger.getAllAppenders.hasMoreElements
       if (!log4j12Initialized) {
-        val defaultLogProps = "org/apache/spark/log4j-defaults.properties"
-        Option(Utils.getSparkClassLoader.getResource(defaultLogProps)) match {
-          case Some(url) =>
-            PropertyConfigurator.configure(url)
-            System.err.println(s"Using Spark's default log4j profile: $defaultLogProps")
-          case None =>
-            System.err.println(s"Spark was unable to load $defaultLogProps")
+        if (Utils.isInInterpreter) {
+          val replDefaultLogProps = "org/apache/spark/log4j-defaults-repl.properties"
+          Option(Utils.getSparkClassLoader.getResource(replDefaultLogProps)) match {
+            case Some(url) =>
+              PropertyConfigurator.configure(url)
+              System.err.println(s"Using Spark's repl log4j profile: $replDefaultLogProps")
+              System.err.println("To adjust logging level use sc.setLogLevel(\"INFO\")")
+            case None =>
+              System.err.println(s"Spark was unable to load $replDefaultLogProps")
+          }
+        } else {
+          val defaultLogProps = "org/apache/spark/log4j-defaults.properties"
+          Option(Utils.getSparkClassLoader.getResource(defaultLogProps)) match {
+            case Some(url) =>
+              PropertyConfigurator.configure(url)
+              System.err.println(s"Using Spark's default log4j profile: $defaultLogProps")
+            case None =>
+              System.err.println(s"Spark was unable to load $defaultLogProps")
+          }
         }
       }
     }

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

@@ -21,7 +21,7 @@ import java.io._
 import java.util.concurrent.ConcurrentHashMap
 import java.util.zip.{GZIPInputStream, GZIPOutputStream}
 
-import scala.collection.mutable.{HashSet, Map}
+import scala.collection.mutable.{HashMap, HashSet, Map}
 import scala.collection.JavaConversions._
 import scala.reflect.ClassTag
 
@@ -284,6 +284,53 @@ private[spark] class MapOutputTrackerMaster(conf: SparkConf)
     cachedSerializedStatuses.contains(shuffleId) || mapStatuses.contains(shuffleId)
   }
 
+  /**
+   * Return a list of locations that each have fraction of map output greater than the specified
+   * threshold.
+   *
+   * @param shuffleId id of the shuffle
+   * @param reducerId id of the reduce task
+   * @param numReducers total number of reducers in the shuffle
+   * @param fractionThreshold fraction of total map output size that a location must have
+   *                          for it to be considered large.
+   *
+   * This method is not thread-safe.
+   */
+  def getLocationsWithLargestOutputs(
+      shuffleId: Int,
+      reducerId: Int,
+      numReducers: Int,
+      fractionThreshold: Double)
+    : Option[Array[BlockManagerId]] = {
+
+    if (mapStatuses.contains(shuffleId)) {
+      val statuses = mapStatuses(shuffleId)
+      if (statuses.nonEmpty) {
+        // HashMap to add up sizes of all blocks at the same location
+        val locs = new HashMap[BlockManagerId, Long]
+        var totalOutputSize = 0L
+        var mapIdx = 0
+        while (mapIdx < statuses.length) {
+          val status = statuses(mapIdx)
+          val blockSize = status.getSizeForBlock(reducerId)
+          if (blockSize > 0) {
+            locs(status.location) = locs.getOrElse(status.location, 0L) + blockSize
+            totalOutputSize += blockSize
+          }
+          mapIdx = mapIdx + 1
+        }
+        val topLocs = locs.filter { case (loc, size) =>
+          size.toDouble / totalOutputSize >= fractionThreshold
+        }
+        // Return if we have any locations which satisfy the required threshold
+        if (topLocs.nonEmpty) {
+          return Some(topLocs.map(_._1).toArray)
+        }
+      }
+    }
+    None
+  }
+
   def incrementEpoch() {
     epochLock.synchronized {
       epoch += 1

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

@@ -20,6 +20,8 @@ package org.apache.spark
 import java.io.File
 import java.net.Socket
 
+import akka.actor.ActorSystem
+
 import scala.collection.JavaConversions._
 import scala.collection.mutable
 import scala.util.Properties
@@ -75,7 +77,8 @@ class SparkEnv (
     val conf: SparkConf) extends Logging {
 
   // TODO Remove actorSystem
-  val actorSystem = rpcEnv.asInstanceOf[AkkaRpcEnv].actorSystem
+  @deprecated("Actor system is no longer supported as of 1.4")
+  val actorSystem: ActorSystem = rpcEnv.asInstanceOf[AkkaRpcEnv].actorSystem
 
   private[spark] var isStopped = false
   private val pythonWorkers = mutable.HashMap[(String, Map[String, String]), PythonWorkerFactory]()

core/src/main/scala/org/apache/spark/api/r/RBackend.scala

@@ -29,7 +29,7 @@ import io.netty.channel.socket.nio.NioServerSocketChannel
 import io.netty.handler.codec.LengthFieldBasedFrameDecoder
 import io.netty.handler.codec.bytes.{ByteArrayDecoder, ByteArrayEncoder}
 
-import org.apache.spark.Logging
+import org.apache.spark.{Logging, SparkConf}
 
 /**
  * Netty-based backend server that is used to communicate between R and Java.
@@ -41,7 +41,8 @@ private[spark] class RBackend {
   private[this] var bossGroup: EventLoopGroup = null
 
   def init(): Int = {
-    bossGroup = new NioEventLoopGroup(2)
+    val conf = new SparkConf()
+    bossGroup = new NioEventLoopGroup(conf.getInt("spark.r.numRBackendThreads", 2))
     val workerGroup = bossGroup
     val handler = new RBackendHandler(this)
 

core/src/main/scala/org/apache/spark/api/r/SerDe.scala

@@ -18,7 +18,7 @@
 package org.apache.spark.api.r
 
 import java.io.{DataInputStream, DataOutputStream}
-import java.sql.{Date, Time}
+import java.sql.{Timestamp, Date, Time}
 
 import scala.collection.JavaConversions._
 
@@ -107,9 +107,12 @@ private[spark] object SerDe {
     Date.valueOf(readString(in))
   }
 
-  def readTime(in: DataInputStream): Time = {
-    val t = in.readDouble()
-    new Time((t * 1000L).toLong)
+  def readTime(in: DataInputStream): Timestamp = {
+    val seconds = in.readDouble()
+    val sec = Math.floor(seconds).toLong
+    val t = new Timestamp(sec * 1000L)
+    t.setNanos(((seconds - sec) * 1e9).toInt)
+    t
   }
 
   def readBytesArr(in: DataInputStream): Array[Array[Byte]] = {
@@ -227,6 +230,9 @@ private[spark] object SerDe {
         case "java.sql.Time" =>
           writeType(dos, "time")
           writeTime(dos, value.asInstanceOf[Time])
+        case "java.sql.Timestamp" =>
+          writeType(dos, "time")
+          writeTime(dos, value.asInstanceOf[Timestamp])
         case "[B" =>
           writeType(dos, "raw")
           writeBytes(dos, value.asInstanceOf[Array[Byte]])
@@ -289,6 +295,9 @@ private[spark] object SerDe {
     out.writeDouble(value.getTime.toDouble / 1000.0)
   }
 
+  def writeTime(out: DataOutputStream, value: Timestamp): Unit = {
+    out.writeDouble((value.getTime / 1000).toDouble + value.getNanos.toDouble / 1e9)
+  }
 
   // NOTE: Only works for ASCII right now
   def writeString(out: DataOutputStream, value: String): Unit = {

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

@@ -324,55 +324,20 @@ object SparkSubmit {
         // Usage: PythonAppRunner <main python file> <extra python files> [app arguments]
         args.mainClass = "org.apache.spark.deploy.PythonRunner"
         args.childArgs = ArrayBuffer(args.primaryResource, args.pyFiles) ++ args.childArgs
-        args.files = mergeFileLists(args.files, args.primaryResource)
+        if (clusterManager != YARN) {
+          // The YARN backend distributes the primary file differently, so don't merge it.
+          args.files = mergeFileLists(args.files, args.primaryResource)
+        }
+      }
+      if (clusterManager != YARN) {
+        // The YARN backend handles python files differently, so don't merge the lists.
+        args.files = mergeFileLists(args.files, args.pyFiles)
       }
-      args.files = mergeFileLists(args.files, args.pyFiles)
       if (args.pyFiles != null) {
         sysProps("spark.submit.pyFiles") = args.pyFiles
       }
     }
 
-    // In yarn mode for a python app, add pyspark archives to files
-    // that can be distributed with the job
-    if (args.isPython && clusterManager == YARN) {
-      var pyArchives: String = null
-      val pyArchivesEnvOpt = sys.env.get("PYSPARK_ARCHIVES_PATH")
-      if (pyArchivesEnvOpt.isDefined) {
-        pyArchives = pyArchivesEnvOpt.get
-      } else {
-        if (!sys.env.contains("SPARK_HOME")) {
-          printErrorAndExit("SPARK_HOME does not exist for python application in yarn mode.")
-        }
-        val pythonPath = new ArrayBuffer[String]
-        for (sparkHome <- sys.env.get("SPARK_HOME")) {
-          val pyLibPath = Seq(sparkHome, "python", "lib").mkString(File.separator)
-          val pyArchivesFile = new File(pyLibPath, "pyspark.zip")
-          if (!pyArchivesFile.exists()) {
-            printErrorAndExit("pyspark.zip does not exist for python application in yarn mode.")
-          }
-          val py4jFile = new File(pyLibPath, "py4j-0.8.2.1-src.zip")
-          if (!py4jFile.exists()) {
-            printErrorAndExit("py4j-0.8.2.1-src.zip does not exist for python application " +
-              "in yarn mode.")
-          }
-          pythonPath += pyArchivesFile.getAbsolutePath()
-          pythonPath += py4jFile.getAbsolutePath()
-        }
-        pyArchives = pythonPath.mkString(",")
-      }
-
-      pyArchives = pyArchives.split(",").map { localPath =>
-        val localURI = Utils.resolveURI(localPath)
-        if (localURI.getScheme != "local") {
-          args.files = mergeFileLists(args.files, localURI.toString)
-          new Path(localPath).getName
-        } else {
-          localURI.getPath
-        }
-      }.mkString(File.pathSeparator)
-      sysProps("spark.submit.pyArchives") = pyArchives
-    }
-
     // If we're running a R app, set the main class to our specific R runner
     if (args.isR && deployMode == CLIENT) {
       if (args.primaryResource == SPARKR_SHELL) {
@@ -386,19 +351,10 @@ object SparkSubmit {
       }
     }
 
-    if (isYarnCluster) {
-      // In yarn-cluster mode for a python app, add primary resource and pyFiles to files
-      // that can be distributed with the job
-      if (args.isPython) {
-        args.files = mergeFileLists(args.files, args.primaryResource)
-        args.files = mergeFileLists(args.files, args.pyFiles)
-      }
-
+    if (isYarnCluster && args.isR) {
       // In yarn-cluster mode for a R app, add primary resource to files
       // that can be distributed with the job
-      if (args.isR) {
-        args.files = mergeFileLists(args.files, args.primaryResource)
-      }
+      args.files = mergeFileLists(args.files, args.primaryResource)
     }
 
     // Special flag to avoid deprecation warnings at the client
@@ -515,17 +471,18 @@ object SparkSubmit {
       }
     }
 
+    // Let YARN know it's a pyspark app, so it distributes needed libraries.
+    if (clusterManager == YARN && args.isPython) {
+      sysProps.put("spark.yarn.isPython", "true")
+    }
+
     // In yarn-cluster mode, use yarn.Client as a wrapper around the user class
     if (isYarnCluster) {
       childMainClass = "org.apache.spark.deploy.yarn.Client"
       if (args.isPython) {
-        val mainPyFile = new Path(args.primaryResource).getName
-        childArgs += ("--primary-py-file", mainPyFile)
+        childArgs += ("--primary-py-file", args.primaryResource)
         if (args.pyFiles != null) {
-          // These files will be distributed to each machine's working directory, so strip the
-          // path prefix
-          val pyFilesNames = args.pyFiles.split(",").map(p => (new Path(p)).getName).mkString(",")
-          childArgs += ("--py-files", pyFilesNames)
+          childArgs += ("--py-files", args.pyFiles)
         }
         childArgs += ("--class", "org.apache.spark.deploy.PythonRunner")
       } else if (args.isR) {

core/src/main/scala/org/apache/spark/rdd/RDDOperationScope.scala

@@ -95,10 +95,9 @@ private[spark] object RDDOperationScope extends Logging {
   private[spark] def withScope[T](
       sc: SparkContext,
       allowNesting: Boolean = false)(body: => T): T = {
-    val stackTrace = Thread.currentThread.getStackTrace().tail // ignore "Thread#getStackTrace"
-    val ourMethodName = stackTrace(1).getMethodName // i.e. withScope
-    // Climb upwards to find the first method that's called something different
-    val callerMethodName = stackTrace
+    val ourMethodName = "withScope"
+    val callerMethodName = Thread.currentThread.getStackTrace()
+      .dropWhile(_.getMethodName != ourMethodName)
       .find(_.getMethodName != ourMethodName)
       .map(_.getMethodName)
       .getOrElse {