Partition previous state RDD if partitioner not present

streaming/src/main/scala/org/apache/spark/streaming/dstream/TrackStateDStream.scala

@@ -132,22 +132,39 @@ class InternalTrackStateDStream[K: ClassTag, V: ClassTag, S: ClassTag, E: ClassT
   /** Method that generates a RDD for the given time */
   override def compute(validTime: Time): Option[RDD[TrackStateRDDRecord[K, S, E]]] = {
     // Get the previous state or create a new empty state RDD
-    val prevStateRDD = getOrCompute(validTime - slideDuration).getOrElse {
-      TrackStateRDD.createFromPairRDD[K, V, S, E](
-        spec.getInitialStateRDD().getOrElse(new EmptyRDD[(K, S)](ssc.sparkContext)),
-        partitioner, validTime
-      )
+    val prevStateRDD = getOrCompute(validTime - slideDuration) match {
+      case Some(rdd) =>
+        if (rdd.partitioner != Some(partitioner)) {
+          // If the RDD is not partitioned the right way, let us repartition it using the
+          // partition index as the key. This is to ensure that state RDD is always partitioned
+          // before creating another state RDD using it
+          val kvRDD = rdd.mapPartitions { iter =>
+            iter.map { x => (TaskContext.get().partitionId(), x)}
+          }
+          kvRDD.partitionBy(partitioner).mapPartitions(iter => iter.map { _._2 },
+            preservesPartitioning = true)
+        } else {
+          rdd
+        }
+      case None =>
+        TrackStateRDD.createFromPairRDD[K, V, S, E](
+          spec.getInitialStateRDD().getOrElse(new EmptyRDD[(K, S)](ssc.sparkContext)),
+          partitioner, validTime
+        )
     }
 
+
     // Compute the new state RDD with previous state RDD and partitioned data RDD
-    parent.getOrCompute(validTime).map { dataRDD =>
-      val partitionedDataRDD = dataRDD.partitionBy(partitioner)
-      val timeoutThresholdTime = spec.getTimeoutInterval().map { interval =>
-        (validTime - interval).milliseconds
-      }
-      new TrackStateRDD(
-        prevStateRDD, partitionedDataRDD, trackingFunction, validTime, timeoutThresholdTime)
+    // Even if there is no data RDD, use an empty one to create a new state RDD
+    val dataRDD = parent.getOrCompute(validTime).getOrElse {
+      context.sparkContext.emptyRDD[(K, V)]
+    }
+    val partitionedDataRDD = dataRDD.partitionBy(partitioner)
+    val timeoutThresholdTime = spec.getTimeoutInterval().map { interval =>
+      (validTime - interval).milliseconds
     }
+    Some(new TrackStateRDD(
+      prevStateRDD, partitionedDataRDD, trackingFunction, validTime, timeoutThresholdTime))
   }
 }
 

streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala

@@ -33,17 +33,133 @@ import org.mockito.Mockito.mock
 import org.scalatest.concurrent.Eventually._
 import org.scalatest.time.SpanSugar._
 
-import org.apache.spark.TestUtils
+import org.apache.spark.{SparkConf, SparkContext, SparkFunSuite, TestUtils}
 import org.apache.spark.streaming.dstream.{DStream, FileInputDStream}
 import org.apache.spark.streaming.scheduler._
 import org.apache.spark.util.{MutableURLClassLoader, Clock, ManualClock, Utils}
 
+/**
+ * A trait of that can be mixed in to get methods for testing DStream operations under
+ * DStream checkpointing. Note that the implementations of this trait has to implement
+ * the `setupCheckpointOperation`
+ */
+trait DStreamCheckpointTester { self: SparkFunSuite =>
+
+  /**
+   * Tests a streaming operation under checkpointing, by restarting the operation
+   * from checkpoint file and verifying whether the final output is correct.
+   * The output is assumed to have come from a reliable queue which an replay
+   * data as required.
+   *
+   * NOTE: This takes into consideration that the last batch processed before
+   * master failure will be re-processed after restart/recovery.
+   */
+  protected def testCheckpointedOperation[U: ClassTag, V: ClassTag](
+      input: Seq[Seq[U]],
+      operation: DStream[U] => DStream[V],
+      expectedOutput: Seq[Seq[V]],
+      numBatchesBeforeRestart: Int,
+      batchDuration: Duration = Seconds(1),
+      stopSparkContextAfterTest: Boolean = true
+    ) {
+    require(numBatchesBeforeRestart < expectedOutput.size,
+      "Number of batches before context restart less than number of expected output " +
+        "(i.e. number of total batches to run)")
+    require(StreamingContext.getActive().isEmpty,
+      "Cannot run test with already active streaming context")
+
+    // Current code assumes that:
+    // number of inputs = number of outputs = number of batches to be run
+    val totalNumBatches = input.size
+    val nextNumBatches = totalNumBatches - numBatchesBeforeRestart
+    val initialNumExpectedOutputs = numBatchesBeforeRestart
+    val nextNumExpectedOutputs = expectedOutput.size - initialNumExpectedOutputs + 1
+    // because the last batch will be processed again
+
+    // Setup the stream computation
+    val checkpointDir = Utils.createTempDir(this.getClass.getSimpleName()).toString
+    logDebug(s"Using checkpoint directory $checkpointDir")
+    val ssc = createContextForCheckpointOperation(batchDuration)
+    require(ssc.conf.get("spark.streaming.clock") === classOf[ManualClock].getName,
+      "Cannot run test without manual clock in the conf")
+
+    val inputStream = new TestInputStream(ssc, input, numPartitions = 2)
+    val operatedStream = operation(inputStream)
+    val outputStream = new TestOutputStreamWithPartitions(operatedStream,
+      new ArrayBuffer[Seq[Seq[V]]] with SynchronizedBuffer[Seq[Seq[V]]])
+    outputStream.register()
+    ssc.checkpoint(checkpointDir)
+
+    // Do the computation for initial number of batches, create checkpoint file and quit
+    generateAndAssertOutput[V](ssc, batchDuration, checkpointDir,
+      expectedOutput.take(numBatchesBeforeRestart), stopSparkContextAfterTest)
+
+    // Restart and complete the computation from checkpoint file
+    logInfo(
+      "\n-------------------------------------------\n" +
+        "        Restarting stream computation          " +
+        "\n-------------------------------------------\n"
+    )
+    val restartedSsc = new StreamingContext(checkpointDir)
+    generateAndAssertOutput[V](restartedSsc, batchDuration, checkpointDir,
+      expectedOutput.takeRight(nextNumExpectedOutputs), stopSparkContextAfterTest)
+  }
+
+  protected def createContextForCheckpointOperation(batchDuration: Duration): StreamingContext = {
+    val conf = new SparkConf().setMaster("local").setAppName(this.getClass.getSimpleName)
+    conf.set("spark.streaming.clock", classOf[ManualClock].getName())
+    new StreamingContext(SparkContext.getOrCreate(conf), Seconds(1))
+  }
+
+  private def generateAndAssertOutput[V: ClassTag](
+      ssc: StreamingContext,
+      batchDuration: Duration,
+      checkpointDir: String,
+      expectedOutput: Seq[Seq[V]],
+      stopSparkContext: Boolean
+    ) {
+    try {
+      ssc.start()
+      val numBatches = expectedOutput.size
+      val clock = ssc.scheduler.clock.asInstanceOf[ManualClock]
+      logDebug("Manual clock before advancing = " + clock.getTimeMillis())
+      clock.advance((batchDuration * numBatches).milliseconds)
+      logDebug("Manual clock after advancing = " + clock.getTimeMillis())
+
+      val outputStream = ssc.graph.getOutputStreams().filter { dstream =>
+        dstream.isInstanceOf[TestOutputStreamWithPartitions[V]]
+      }.head.asInstanceOf[TestOutputStreamWithPartitions[V]]
+
+      eventually(timeout(10 seconds)) {
+        ssc.awaitTerminationOrTimeout(10)
+        assert(outputStream.output.size === expectedOutput.size)
+      }
+
+      eventually(timeout(10 seconds)) {
+        Checkpoint.getCheckpointFiles(checkpointDir).exists {
+          _.toString.contains(clock.getTimeMillis.toString)
+        }
+      }
+
+      val output = outputStream.output.map(_.flatten)
+      assert(
+        output.zip(expectedOutput).forall { case (o, e) => o.toSet === e.toSet },
+        s"Set comparison failed\n" +
+          s"Expected output (${expectedOutput.size} items):\n${expectedOutput.mkString("\n")}\n" +
+          s"Generated output (${output.size} items): ${output.mkString("\n")}"
+      )
+    } finally {
+      ssc.stop(stopSparkContext = stopSparkContext)
+    }
+  }
+}
+
 /**
  * This test suites tests the checkpointing functionality of DStreams -
  * the checkpointing of a DStream's RDDs as well as the checkpointing of
  * the whole DStream graph.
  */
-class CheckpointSuite extends TestSuiteBase {
+class CheckpointSuite extends TestSuiteBase with DStreamCheckpointTester {
 
   var ssc: StreamingContext = null
 
@@ -56,7 +172,7 @@ class CheckpointSuite extends TestSuiteBase {
 
   override def afterFunction() {
     super.afterFunction()
-    if (ssc != null) ssc.stop()
+    StreamingContext.getActive().foreach { _.stop() }
     Utils.deleteRecursively(new File(checkpointDir))
   }
 
@@ -634,53 +750,6 @@ class CheckpointSuite extends TestSuiteBase {
     checkpointWriter.stop()
   }
 
-  /**
-   * Tests a streaming operation under checkpointing, by restarting the operation
-   * from checkpoint file and verifying whether the final output is correct.
-   * The output is assumed to have come from a reliable queue which an replay
-   * data as required.
-   *
-   * NOTE: This takes into consideration that the last batch processed before
-   * master failure will be re-processed after restart/recovery.
-   */
-  def testCheckpointedOperation[U: ClassTag, V: ClassTag](
-    input: Seq[Seq[U]],
-    operation: DStream[U] => DStream[V],
-    expectedOutput: Seq[Seq[V]],
-    initialNumBatches: Int
-  ) {
-
-    // Current code assumes that:
-    // number of inputs = number of outputs = number of batches to be run
-    val totalNumBatches = input.size
-    val nextNumBatches = totalNumBatches - initialNumBatches
-    val initialNumExpectedOutputs = initialNumBatches
-    val nextNumExpectedOutputs = expectedOutput.size - initialNumExpectedOutputs + 1
-    // because the last batch will be processed again
-
-    // Do the computation for initial number of batches, create checkpoint file and quit
-    ssc = setupStreams[U, V](input, operation)
-    ssc.start()
-    val output = advanceTimeWithRealDelay[V](ssc, initialNumBatches)
-    ssc.stop()
-    verifyOutput[V](output, expectedOutput.take(initialNumBatches), true)
-    Thread.sleep(1000)
-
-    // Restart and complete the computation from checkpoint file
-    logInfo(
-      "\n-------------------------------------------\n" +
-      "        Restarting stream computation          " +
-      "\n-------------------------------------------\n"
-    )
-    ssc = new StreamingContext(checkpointDir)
-    ssc.start()
-    val outputNew = advanceTimeWithRealDelay[V](ssc, nextNumBatches)
-    // the first element will be re-processed data of the last batch before restart
-    verifyOutput[V](outputNew, expectedOutput.takeRight(nextNumExpectedOutputs), true)
-    ssc.stop()
-    ssc = null
-  }
-
   /**
    * Advances the manual clock on the streaming scheduler by given number of batches.
    * It also waits for the expected amount of time for each batch.

streaming/src/test/scala/org/apache/spark/streaming/TrackStateByKeySuite.scala

@@ -25,31 +25,27 @@ import scala.reflect.ClassTag
 import org.scalatest.PrivateMethodTester._
 import org.scalatest.{BeforeAndAfter, BeforeAndAfterAll}
 
-import org.apache.spark.streaming.dstream.{InternalTrackStateDStream, TrackStateDStream, TrackStateDStreamImpl}
+import org.apache.spark.streaming.dstream.{DStream, InternalTrackStateDStream, TrackStateDStream, TrackStateDStreamImpl}
 import org.apache.spark.util.{ManualClock, Utils}
 import org.apache.spark.{SparkConf, SparkContext, SparkFunSuite}
 
-class TrackStateByKeySuite extends SparkFunSuite with BeforeAndAfterAll with BeforeAndAfter {
+class TrackStateByKeySuite extends SparkFunSuite
+  with DStreamCheckpointTester with BeforeAndAfterAll with BeforeAndAfter {
 
   private var sc: SparkContext = null
-  private var ssc: StreamingContext = null
-  private var checkpointDir: File = null
-  private val batchDuration = Seconds(1)
+  protected var checkpointDir: File = null
+  protected val batchDuration = Seconds(1)
 
   before {
-    StreamingContext.getActive().foreach {
-      _.stop(stopSparkContext = false)
-    }
+    StreamingContext.getActive().foreach { _.stop(stopSparkContext = false) }
     checkpointDir = Utils.createTempDir("checkpoint")
-
-    ssc = new StreamingContext(sc, batchDuration)
-    ssc.checkpoint(checkpointDir.toString)
   }
 
   after {
-    StreamingContext.getActive().foreach {
-      _.stop(stopSparkContext = false)
+    if (checkpointDir != null) {
+      Utils.deleteRecursively(checkpointDir)
     }
+    StreamingContext.getActive().foreach { _.stop(stopSparkContext = false) }
   }
 
   override def beforeAll(): Unit = {
@@ -242,7 +238,7 @@ class TrackStateByKeySuite extends SparkFunSuite with BeforeAndAfterAll with Bef
       assert(dstreamImpl.stateClass === classOf[Double])
       assert(dstreamImpl.emittedClass === classOf[Long])
     }
-
+    val ssc = new StreamingContext(sc, batchDuration)
     val inputStream = new TestInputStream[(String, Int)](ssc, Seq.empty, numPartitions = 2)
 
     // Defining StateSpec inline with trackStateByKey and simple function implicitly gets the types
@@ -451,8 +447,9 @@ class TrackStateByKeySuite extends SparkFunSuite with BeforeAndAfterAll with Bef
         expectedCheckpointDuration: Duration,
         explicitCheckpointDuration: Option[Duration] = None
       ): Unit = {
+      val ssc = new StreamingContext(sc, batchDuration)
+
       try {
-        ssc = new StreamingContext(sc, batchDuration)
         val inputStream = new TestInputStream(ssc, Seq.empty[Seq[Int]], 2).map(_ -> 1)
         val dummyFunc = (value: Option[Int], state: State[Int]) => 0
         val trackStateStream = inputStream.trackStateByKey(StateSpec.function(dummyFunc))
@@ -462,11 +459,12 @@ class TrackStateByKeySuite extends SparkFunSuite with BeforeAndAfterAll with Bef
           trackStateStream.checkpoint(d)
         }
         trackStateStream.register()
+        ssc.checkpoint(checkpointDir.toString)
         ssc.start()  // should initialize all the checkpoint durations
         assert(trackStateStream.checkpointDuration === null)
         assert(internalTrackStateStream.checkpointDuration === expectedCheckpointDuration)
       } finally {
-        StreamingContext.getActive().foreach { _.stop(stopSparkContext = false) }
+        ssc.stop(stopSparkContext = false)
       }
     }
 
@@ -479,6 +477,50 @@ class TrackStateByKeySuite extends SparkFunSuite with BeforeAndAfterAll with Bef
     testCheckpointDuration(Seconds(10), Seconds(20), Some(Seconds(20)))
   }
 
+
+  test("trackStateByKey - drivery failure recovery") {
+    val inputData =
+      Seq(
+        Seq(),
+        Seq("a"),
+        Seq("a", "b"),
+        Seq("a", "b", "c"),
+        Seq("a", "b"),
+        Seq("a"),
+        Seq()
+      )
+
+    val stateData =
+      Seq(
+        Seq(),
+        Seq(("a", 1)),
+        Seq(("a", 2), ("b", 1)),
+        Seq(("a", 3), ("b", 2), ("c", 1)),
+        Seq(("a", 4), ("b", 3), ("c", 1)),
+        Seq(("a", 5), ("b", 3), ("c", 1)),
+        Seq(("a", 5), ("b", 3), ("c", 1))
+      )
+
+    def operation(dstream: DStream[String]): DStream[(String, Int)] = {
+
+      val checkpointDuration = batchDuration * (stateData.size / 2)
+
+      val runningCount = (value: Option[Int], state: State[Int]) => {
+        state.update(state.getOption().getOrElse(0) + value.getOrElse(0))
+        state.get()
+      }
+
+      val trackStateStream = dstream.map { _ -> 1 }.trackStateByKey(
+        StateSpec.function(runningCount))
+      // Set internval make sure there is one RDD checkpointing
+      trackStateStream.checkpoint(checkpointDuration)
+      trackStateStream.stateSnapshots()
+    }
+
+    testCheckpointedOperation(inputData, operation, stateData, inputData.size / 2,
+      batchDuration = batchDuration, stopSparkContextAfterTest = false)
+  }
+
   private def testOperation[K: ClassTag, S: ClassTag, T: ClassTag](
       input: Seq[Seq[K]],
       trackStateSpec: StateSpec[K, Int, S, T],
@@ -500,6 +542,7 @@ class TrackStateByKeySuite extends SparkFunSuite with BeforeAndAfterAll with Bef
     ): (Seq[Seq[T]], Seq[Seq[(K, S)]]) = {
 
     // Setup the stream computation
+    val ssc = new StreamingContext(sc, Seconds(1))
     val inputStream = new TestInputStream(ssc, input, numPartitions = 2)
     val trackeStateStream = inputStream.map(x => (x, 1)).trackStateByKey(trackStateSpec)
     val collectedOutputs = new ArrayBuffer[Seq[T]] with SynchronizedBuffer[Seq[T]]
@@ -511,12 +554,14 @@ class TrackStateByKeySuite extends SparkFunSuite with BeforeAndAfterAll with Bef
     stateSnapshotStream.register()
 
     val batchCounter = new BatchCounter(ssc)
+    ssc.checkpoint(checkpointDir.toString)
     ssc.start()
 
     val clock = ssc.scheduler.clock.asInstanceOf[ManualClock]
     clock.advance(batchDuration.milliseconds * numBatches)
 
     batchCounter.waitUntilBatchesCompleted(numBatches, 10000)
+    ssc.stop(stopSparkContext = false)
     (collectedOutputs, collectedStateSnapshots)
   }