diff options
Diffstat (limited to 'streaming')
9 files changed, 180 insertions, 15 deletions
diff --git a/streaming/pom.xml b/streaming/pom.xml index 8022c4fe18..7a9ae6a97b 100644 --- a/streaming/pom.xml +++ b/streaming/pom.xml @@ -73,6 +73,10 @@ <groupId>com.sun.jdmk</groupId> <artifactId>jmxtools</artifactId> </exclusion> + <exclusion> + <groupId>net.sf.jopt-simple</groupId> + <artifactId>jopt-simple</artifactId> + </exclusion> </exclusions> </dependency> <dependency> diff --git a/streaming/src/main/scala/org/apache/spark/streaming/DStream.scala b/streaming/src/main/scala/org/apache/spark/streaming/DStream.scala index 38e34795b4..9ceff754c4 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/DStream.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/DStream.scala @@ -438,6 +438,13 @@ abstract class DStream[T: ClassManifest] ( */ def glom(): DStream[Array[T]] = new GlommedDStream(this) + + /** + * Return a new DStream with an increased or decreased level of parallelism. Each RDD in the + * returned DStream has exactly numPartitions partitions. + */ + def repartition(numPartitions: Int): DStream[T] = this.transform(_.repartition(numPartitions)) + /** * Return a new DStream in which each RDD is generated by applying mapPartitions() to each RDDs * of this DStream. Applying mapPartitions() to an RDD applies a function to each partition diff --git a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala index d1932b6b05..1a2aeaa879 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaDStream.scala @@ -94,6 +94,12 @@ class JavaDStream[T](val dstream: DStream[T])(implicit val classManifest: ClassM */ def union(that: JavaDStream[T]): JavaDStream[T] = dstream.union(that.dstream) + + /** + * Return a new DStream with an increased or decreased level of parallelism. Each RDD in the + * returned DStream has exactly numPartitions partitions. + */ + def repartition(numPartitions: Int): JavaDStream[T] = dstream.repartition(numPartitions) } object JavaDStream { diff --git a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala index 4dd6b7d096..c6cd635afa 100644 --- a/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala +++ b/streaming/src/main/scala/org/apache/spark/streaming/api/java/JavaPairDStream.scala @@ -59,6 +59,12 @@ class JavaPairDStream[K, V](val dstream: DStream[(K, V)])( /** Persist the RDDs of this DStream with the given storage level */ def persist(storageLevel: StorageLevel): JavaPairDStream[K, V] = dstream.persist(storageLevel) + /** + * Return a new DStream with an increased or decreased level of parallelism. Each RDD in the + * returned DStream has exactly numPartitions partitions. + */ + def repartition(numPartitions: Int): JavaPairDStream[K, V] = dstream.repartition(numPartitions) + /** Method that generates a RDD for the given Duration */ def compute(validTime: Time): JavaPairRDD[K, V] = { dstream.compute(validTime) match { diff --git a/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java b/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java index 5d48908667..ad4a8b9535 100644 --- a/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java +++ b/streaming/src/test/java/org/apache/spark/streaming/JavaAPISuite.java @@ -185,6 +185,39 @@ public class JavaAPISuite implements Serializable { } @Test + public void testRepartitionMorePartitions() { + List<List<Integer>> inputData = Arrays.asList( + Arrays.asList(1, 2, 3, 4, 5, 6, 7, 8, 9, 10), + Arrays.asList(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)); + JavaDStream stream = JavaTestUtils.attachTestInputStream(ssc, inputData, 2); + JavaDStream repartitioned = stream.repartition(4); + JavaTestUtils.attachTestOutputStream(repartitioned); + List<List<List<Integer>>> result = JavaTestUtils.runStreamsWithPartitions(ssc, 2, 2); + Assert.assertEquals(2, result.size()); + for (List<List<Integer>> rdd : result) { + Assert.assertEquals(4, rdd.size()); + Assert.assertEquals( + 10, rdd.get(0).size() + rdd.get(1).size() + rdd.get(2).size() + rdd.get(3).size()); + } + } + + @Test + public void testRepartitionFewerPartitions() { + List<List<Integer>> inputData = Arrays.asList( + Arrays.asList(1, 2, 3, 4, 5, 6, 7, 8, 9, 10), + Arrays.asList(1, 2, 3, 4, 5, 6, 7, 8, 9, 10)); + JavaDStream stream = JavaTestUtils.attachTestInputStream(ssc, inputData, 4); + JavaDStream repartitioned = stream.repartition(2); + JavaTestUtils.attachTestOutputStream(repartitioned); + List<List<List<Integer>>> result = JavaTestUtils.runStreamsWithPartitions(ssc, 2, 2); + Assert.assertEquals(2, result.size()); + for (List<List<Integer>> rdd : result) { + Assert.assertEquals(2, rdd.size()); + Assert.assertEquals(10, rdd.get(0).size() + rdd.get(1).size()); + } + } + + @Test public void testGlom() { List<List<String>> inputData = Arrays.asList( Arrays.asList("giants", "dodgers"), diff --git a/streaming/src/test/java/org/apache/spark/streaming/JavaTestUtils.scala b/streaming/src/test/java/org/apache/spark/streaming/JavaTestUtils.scala index 8a6604904d..5e384eeee4 100644 --- a/streaming/src/test/java/org/apache/spark/streaming/JavaTestUtils.scala +++ b/streaming/src/test/java/org/apache/spark/streaming/JavaTestUtils.scala @@ -33,9 +33,9 @@ trait JavaTestBase extends TestSuiteBase { * The stream will be derived from the supplied lists of Java objects. **/ def attachTestInputStream[T]( - ssc: JavaStreamingContext, - data: JList[JList[T]], - numPartitions: Int) = { + ssc: JavaStreamingContext, + data: JList[JList[T]], + numPartitions: Int) = { val seqData = data.map(Seq(_:_*)) implicit val cm: ClassManifest[T] = @@ -50,12 +50,11 @@ trait JavaTestBase extends TestSuiteBase { * [[org.apache.spark.streaming.TestOutputStream]]. **/ def attachTestOutputStream[T, This <: JavaDStreamLike[T, This, R], R <: JavaRDDLike[T, R]]( - dstream: JavaDStreamLike[T, This, R]) = + dstream: JavaDStreamLike[T, This, R]) = { implicit val cm: ClassManifest[T] = implicitly[ClassManifest[AnyRef]].asInstanceOf[ClassManifest[T]] - val ostream = new TestOutputStream(dstream.dstream, - new ArrayBuffer[Seq[T]] with SynchronizedBuffer[Seq[T]]) + val ostream = new TestOutputStreamWithPartitions(dstream.dstream) dstream.dstream.ssc.registerOutputStream(ostream) } @@ -63,9 +62,11 @@ trait JavaTestBase extends TestSuiteBase { * Process all registered streams for a numBatches batches, failing if * numExpectedOutput RDD's are not generated. Generated RDD's are collected * and returned, represented as a list for each batch interval. + * + * Returns a list of items for each RDD. */ def runStreams[V]( - ssc: JavaStreamingContext, numBatches: Int, numExpectedOutput: Int): JList[JList[V]] = { + ssc: JavaStreamingContext, numBatches: Int, numExpectedOutput: Int): JList[JList[V]] = { implicit val cm: ClassManifest[V] = implicitly[ClassManifest[AnyRef]].asInstanceOf[ClassManifest[V]] val res = runStreams[V](ssc.ssc, numBatches, numExpectedOutput) @@ -73,6 +74,27 @@ trait JavaTestBase extends TestSuiteBase { res.map(entry => out.append(new ArrayList[V](entry))) out } + + /** + * Process all registered streams for a numBatches batches, failing if + * numExpectedOutput RDD's are not generated. Generated RDD's are collected + * and returned, represented as a list for each batch interval. + * + * Returns a sequence of RDD's. Each RDD is represented as several sequences of items, each + * representing one partition. + */ + def runStreamsWithPartitions[V](ssc: JavaStreamingContext, numBatches: Int, + numExpectedOutput: Int): JList[JList[JList[V]]] = { + implicit val cm: ClassManifest[V] = + implicitly[ClassManifest[AnyRef]].asInstanceOf[ClassManifest[V]] + val res = runStreamsWithPartitions[V](ssc.ssc, numBatches, numExpectedOutput) + val out = new ArrayList[JList[JList[V]]]() + res.map{entry => + val lists = entry.map(new ArrayList[V](_)) + out.append(new ArrayList[JList[V]](lists)) + } + out + } } object JavaTestUtils extends JavaTestBase { diff --git a/streaming/src/test/scala/org/apache/spark/streaming/BasicOperationsSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/BasicOperationsSuite.scala index a2ac510a98..259ef1608c 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/BasicOperationsSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/BasicOperationsSuite.scala @@ -85,6 +85,44 @@ class BasicOperationsSuite extends TestSuiteBase { testOperation(input, operation, output, true) } + test("repartition (more partitions)") { + val input = Seq(1 to 100, 101 to 200, 201 to 300) + val operation = (r: DStream[Int]) => r.repartition(5) + val ssc = setupStreams(input, operation, 2) + val output = runStreamsWithPartitions(ssc, 3, 3) + assert(output.size === 3) + val first = output(0) + val second = output(1) + val third = output(2) + + assert(first.size === 5) + assert(second.size === 5) + assert(third.size === 5) + + assert(first.flatten.toSet === (1 to 100).toSet) + assert(second.flatten.toSet === (101 to 200).toSet) + assert(third.flatten.toSet === (201 to 300).toSet) + } + + test("repartition (fewer partitions)") { + val input = Seq(1 to 100, 101 to 200, 201 to 300) + val operation = (r: DStream[Int]) => r.repartition(2) + val ssc = setupStreams(input, operation, 5) + val output = runStreamsWithPartitions(ssc, 3, 3) + assert(output.size === 3) + val first = output(0) + val second = output(1) + val third = output(2) + + assert(first.size === 2) + assert(second.size === 2) + assert(third.size === 2) + + assert(first.flatten.toSet === (1 to 100).toSet) + assert(second.flatten.toSet === (101 to 200).toSet) + assert(third.flatten.toSet === (201 to 300).toSet) + } + test("groupByKey") { testOperation( Seq( Seq("a", "a", "b"), Seq("", ""), Seq() ), diff --git a/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala b/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala index a327de80b3..beb20831bd 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/CheckpointSuite.scala @@ -366,7 +366,7 @@ class CheckpointSuite extends TestSuiteBase with BeforeAndAfter { logInfo("Manual clock after advancing = " + clock.time) Thread.sleep(batchDuration.milliseconds) - val outputStream = ssc.graph.getOutputStreams.head.asInstanceOf[TestOutputStream[V]] - outputStream.output + val outputStream = ssc.graph.getOutputStreams.head.asInstanceOf[TestOutputStreamWithPartitions[V]] + outputStream.output.map(_.flatten) } } diff --git a/streaming/src/test/scala/org/apache/spark/streaming/TestSuiteBase.scala b/streaming/src/test/scala/org/apache/spark/streaming/TestSuiteBase.scala index 37dd9c4cc6..be140699c2 100644 --- a/streaming/src/test/scala/org/apache/spark/streaming/TestSuiteBase.scala +++ b/streaming/src/test/scala/org/apache/spark/streaming/TestSuiteBase.scala @@ -60,8 +60,11 @@ class TestInputStream[T: ClassManifest](ssc_ : StreamingContext, input: Seq[Seq[ /** * This is a output stream just for the testsuites. All the output is collected into a * ArrayBuffer. This buffer is wiped clean on being restored from checkpoint. + * + * The buffer contains a sequence of RDD's, each containing a sequence of items */ -class TestOutputStream[T: ClassManifest](parent: DStream[T], val output: ArrayBuffer[Seq[T]]) +class TestOutputStream[T: ClassManifest](parent: DStream[T], + val output: ArrayBuffer[Seq[T]] = ArrayBuffer[Seq[T]]()) extends ForEachDStream[T](parent, (rdd: RDD[T], t: Time) => { val collected = rdd.collect() output += collected @@ -76,6 +79,30 @@ class TestOutputStream[T: ClassManifest](parent: DStream[T], val output: ArrayBu } /** + * This is a output stream just for the testsuites. All the output is collected into a + * ArrayBuffer. This buffer is wiped clean on being restored from checkpoint. + * + * The buffer contains a sequence of RDD's, each containing a sequence of partitions, each + * containing a sequence of items. + */ +class TestOutputStreamWithPartitions[T: ClassManifest](parent: DStream[T], + val output: ArrayBuffer[Seq[Seq[T]]] = ArrayBuffer[Seq[Seq[T]]]()) + extends ForEachDStream[T](parent, (rdd: RDD[T], t: Time) => { + val collected = rdd.glom().collect().map(_.toSeq) + output += collected + }) { + + // This is to clear the output buffer every it is read from a checkpoint + @throws(classOf[IOException]) + private def readObject(ois: ObjectInputStream) { + ois.defaultReadObject() + output.clear() + } + + def toTestOutputStream = new TestOutputStream[T](this.parent, this.output.map(_.flatten)) +} + +/** * This is the base trait for Spark Streaming testsuites. This provides basic functionality * to run user-defined set of input on user-defined stream operations, and verify the output. */ @@ -108,7 +135,8 @@ trait TestSuiteBase extends FunSuite with BeforeAndAfter with Logging { */ def setupStreams[U: ClassManifest, V: ClassManifest]( input: Seq[Seq[U]], - operation: DStream[U] => DStream[V] + operation: DStream[U] => DStream[V], + numPartitions: Int = numInputPartitions ): StreamingContext = { // Create StreamingContext @@ -118,9 +146,10 @@ trait TestSuiteBase extends FunSuite with BeforeAndAfter with Logging { } // Setup the stream computation - val inputStream = new TestInputStream(ssc, input, numInputPartitions) + val inputStream = new TestInputStream(ssc, input, numPartitions) val operatedStream = operation(inputStream) - val outputStream = new TestOutputStream(operatedStream, new ArrayBuffer[Seq[V]] with SynchronizedBuffer[Seq[V]]) + val outputStream = new TestOutputStreamWithPartitions(operatedStream, + new ArrayBuffer[Seq[Seq[V]]] with SynchronizedBuffer[Seq[Seq[V]]]) ssc.registerInputStream(inputStream) ssc.registerOutputStream(outputStream) ssc @@ -146,7 +175,8 @@ trait TestSuiteBase extends FunSuite with BeforeAndAfter with Logging { val inputStream1 = new TestInputStream(ssc, input1, numInputPartitions) val inputStream2 = new TestInputStream(ssc, input2, numInputPartitions) val operatedStream = operation(inputStream1, inputStream2) - val outputStream = new TestOutputStream(operatedStream, new ArrayBuffer[Seq[W]] with SynchronizedBuffer[Seq[W]]) + val outputStream = new TestOutputStreamWithPartitions(operatedStream, + new ArrayBuffer[Seq[Seq[W]]] with SynchronizedBuffer[Seq[Seq[W]]]) ssc.registerInputStream(inputStream1) ssc.registerInputStream(inputStream2) ssc.registerOutputStream(outputStream) @@ -157,18 +187,37 @@ trait TestSuiteBase extends FunSuite with BeforeAndAfter with Logging { * Runs the streams set up in `ssc` on manual clock for `numBatches` batches and * returns the collected output. It will wait until `numExpectedOutput` number of * output data has been collected or timeout (set by `maxWaitTimeMillis`) is reached. + * + * Returns a sequence of items for each RDD. */ def runStreams[V: ClassManifest]( ssc: StreamingContext, numBatches: Int, numExpectedOutput: Int ): Seq[Seq[V]] = { + // Flatten each RDD into a single Seq + runStreamsWithPartitions(ssc, numBatches, numExpectedOutput).map(_.flatten.toSeq) + } + + /** + * Runs the streams set up in `ssc` on manual clock for `numBatches` batches and + * returns the collected output. It will wait until `numExpectedOutput` number of + * output data has been collected or timeout (set by `maxWaitTimeMillis`) is reached. + * + * Returns a sequence of RDD's. Each RDD is represented as several sequences of items, each + * representing one partition. + */ + def runStreamsWithPartitions[V: ClassManifest]( + ssc: StreamingContext, + numBatches: Int, + numExpectedOutput: Int + ): Seq[Seq[Seq[V]]] = { assert(numBatches > 0, "Number of batches to run stream computation is zero") assert(numExpectedOutput > 0, "Number of expected outputs after " + numBatches + " is zero") logInfo("numBatches = " + numBatches + ", numExpectedOutput = " + numExpectedOutput) // Get the output buffer - val outputStream = ssc.graph.getOutputStreams.head.asInstanceOf[TestOutputStream[V]] + val outputStream = ssc.graph.getOutputStreams.head.asInstanceOf[TestOutputStreamWithPartitions[V]] val output = outputStream.output try { |