Examples with SparkManager io.cdap.cdap.test.SparkManager used on opensource projects

Example 11 with SparkManager

use of io.cdap.cdap.test.SparkManager in project cdap by caskdata.

the class MessagingAppTestRun method testSparkMessaging.

@Test
public void testSparkMessaging() throws Exception {
    ApplicationManager appManager = deployWithArtifact(NAMESPACE, MessagingApp.class, artifactJar);
    final SparkManager sparkManager = appManager.getSparkManager(MessagingSpark.class.getSimpleName()).start();
    final MessageFetcher fetcher = getMessagingContext().getMessageFetcher();
    final AtomicReference<String> messageId = new AtomicReference<>();
    // Wait for the Spark to create the topic
    final MessagingAdmin messagingAdmin = getMessagingAdmin(NAMESPACE.getNamespace());
    Tasks.waitFor(true, new Callable<Boolean>() {

        @Override
        public Boolean call() throws Exception {
            try {
                messagingAdmin.getTopicProperties(MessagingApp.TOPIC);
                return true;
            } catch (TopicNotFoundException e) {
                return false;
            }
        }
    }, 60, TimeUnit.SECONDS, 100, TimeUnit.MILLISECONDS);
    // This is to verify failed transaction is not publishing anything.
    for (String expected : Arrays.asList("start", "block")) {
        Tasks.waitFor(expected, new Callable<String>() {

            @Override
            public String call() throws Exception {
                try (CloseableIterator<Message> iterator = fetcher.fetch(NAMESPACE.getNamespace(), MessagingApp.TOPIC, 1, messageId.get())) {
                    if (!iterator.hasNext()) {
                        return null;
                    }
                    Message message = iterator.next();
                    messageId.set(message.getId());
                    return message.getPayloadAsString();
                }
            }
        }, 60, TimeUnit.SECONDS, 100, TimeUnit.MILLISECONDS);
    }
    // Publish a control message to unblock the Spark execution
    getMessagingContext().getMessagePublisher().publish(NAMESPACE.getNamespace(), MessagingApp.CONTROL_TOPIC, "go");
    // Expects a result message as "result-15", where 15 is the sum of 1,2,3,4,5
    Tasks.waitFor("result-15", new Callable<String>() {

        @Override
        public String call() throws Exception {
            try (CloseableIterator<Message> iterator = fetcher.fetch(NAMESPACE.getNamespace(), MessagingApp.TOPIC, 1, messageId.get())) {
                if (!iterator.hasNext()) {
                    return null;
                }
                Message message = iterator.next();
                messageId.set(message.getId());
                return message.getPayloadAsString();
            }
        }
    }, 60, TimeUnit.SECONDS, 100, TimeUnit.MILLISECONDS);
    sparkManager.waitForRun(ProgramRunStatus.COMPLETED, 60, TimeUnit.SECONDS);
}

Example 12 with SparkManager

use of io.cdap.cdap.test.SparkManager in project cdap by caskdata.

the class SparkFileSetTestRun method testSparkWithPartitionedFileSet.

private void testSparkWithPartitionedFileSet(ApplicationManager applicationManager, String sparkProgram) throws Exception {
    DataSetManager<PartitionedFileSet> pfsManager = getDataset("pfs");
    PartitionedFileSet pfs = pfsManager.get();
    PartitionOutput partitionOutput = pfs.getPartitionOutput(PartitionKey.builder().addStringField("x", "nn").build());
    Location location = partitionOutput.getLocation();
    prepareFileInput(location);
    partitionOutput.addPartition();
    pfsManager.flush();
    Map<String, String> inputArgs = new HashMap<>();
    PartitionedFileSetArguments.setInputPartitionFilter(inputArgs, PartitionFilter.builder().addRangeCondition("x", "na", "nx").build());
    Map<String, String> outputArgs = new HashMap<>();
    PartitionKey outputKey = PartitionKey.builder().addStringField("x", "xx").build();
    PartitionedFileSetArguments.setOutputPartitionKey(outputArgs, outputKey);
    Map<String, String> args = new HashMap<>();
    args.putAll(RuntimeArguments.addScope(Scope.DATASET, "pfs", inputArgs));
    args.putAll(RuntimeArguments.addScope(Scope.DATASET, "pfs", outputArgs));
    args.put("input", "pfs");
    args.put("output", "pfs");
    SparkManager sparkManager = applicationManager.getSparkManager(sparkProgram).start(args);
    sparkManager.waitForRun(ProgramRunStatus.COMPLETED, 10, TimeUnit.MINUTES);
    pfsManager.flush();
    PartitionDetail partition = pfs.getPartition(outputKey);
    Assert.assertNotNull(partition);
    validateFileOutput(partition.getLocation());
    // Cleanup after test completed
    pfs.dropPartition(partitionOutput.getPartitionKey());
    pfs.dropPartition(partition.getPartitionKey());
    pfsManager.flush();
}

Example 13 with SparkManager

use of io.cdap.cdap.test.SparkManager in project cdap by caskdata.

the class SparkFileSetTestRun method testSparkWithTimePartitionedFileSet.

private void testSparkWithTimePartitionedFileSet(ApplicationManager applicationManager, String sparkProgram) throws Exception {
    long customOutputPartitionKey = 123456789L;
    long customInputPartitionKey = 987654321L;
    DataSetManager<TimePartitionedFileSet> tpfsManager = getDataset("tpfs");
    long inputTime = System.currentTimeMillis();
    long outputTime = inputTime + TimeUnit.HOURS.toMillis(1);
    addTimePartition(tpfsManager, inputTime);
    addTimePartition(tpfsManager, customInputPartitionKey);
    Map<String, String> inputArgs = new HashMap<>();
    TimePartitionedFileSetArguments.setInputStartTime(inputArgs, inputTime - 100);
    TimePartitionedFileSetArguments.setInputEndTime(inputArgs, inputTime + 100);
    Map<String, String> outputArgs = new HashMap<>();
    TimePartitionedFileSetArguments.setOutputPartitionTime(outputArgs, outputTime);
    Map<String, String> args = new HashMap<>();
    args.putAll(RuntimeArguments.addScope(Scope.DATASET, "tpfs", inputArgs));
    args.putAll(RuntimeArguments.addScope(Scope.DATASET, "tpfs", outputArgs));
    args.put("input", "tpfs");
    args.put("output", "tpfs");
    args.put("outputKey", String.valueOf(customOutputPartitionKey));
    args.put("inputKey", String.valueOf(customInputPartitionKey));
    SparkManager sparkManager = applicationManager.getSparkManager(sparkProgram).start(args);
    sparkManager.waitForRun(ProgramRunStatus.COMPLETED, 10, TimeUnit.MINUTES);
    tpfsManager.flush();
    TimePartitionedFileSet tpfs = tpfsManager.get();
    PartitionDetail partition = tpfs.getPartitionByTime(outputTime);
    Assert.assertNotNull("Output partition is null while for running without custom dataset arguments", partition);
    validateFileOutput(partition.getLocation());
    PartitionDetail customPartition = tpfs.getPartitionByTime(customOutputPartitionKey);
    Assert.assertNotNull("Output partition is null while for running with custom dataset arguments", customPartition);
    validateFileOutput(customPartition.getLocation());
    // Cleanup after running the test
    tpfs.dropPartition(inputTime);
    tpfs.dropPartition(customInputPartitionKey);
    tpfs.dropPartition(partition.getPartitionKey());
    tpfs.dropPartition(customPartition.getPartitionKey());
    tpfsManager.flush();
}

Example 14 with SparkManager

use of io.cdap.cdap.test.SparkManager in project cdap by caskdata.

the class SparkTest method testClassicSpark.

@Test
public void testClassicSpark() throws Exception {
    ApplicationManager appManager = deploy(TestSparkApp.class);
    for (Class<?> sparkClass : Arrays.asList(TestSparkApp.ClassicSpark.class, TestSparkApp.ScalaClassicSpark.class)) {
        final SparkManager sparkManager = appManager.getSparkManager(sparkClass.getSimpleName());
        sparkManager.startAndWaitForGoodRun(ProgramRunStatus.COMPLETED, 5, TimeUnit.MINUTES);
    }
    KeyValueTable resultTable = this.<KeyValueTable>getDataset("ResultTable").get();
    Assert.assertEquals(1L, Bytes.toLong(resultTable.read(ClassicSparkProgram.class.getName())));
    Assert.assertEquals(1L, Bytes.toLong(resultTable.read(ScalaClassicSparkProgram.class.getName())));
}

Example 15 with SparkManager

use of io.cdap.cdap.test.SparkManager in project cdap by caskdata.

the class SparkTest method testTransaction.

@Test
public void testTransaction() throws Exception {
    ApplicationManager applicationManager = deploy(TestSparkApp.class);
    // Write some data to a local file
    File inputFile = TEMP_FOLDER.newFile();
    try (PrintWriter writer = new PrintWriter(Files.newBufferedWriter(inputFile.toPath(), StandardCharsets.UTF_8))) {
        writer.println("red fox");
        writer.println("brown fox");
        writer.println("grey fox");
        writer.println("brown bear");
        writer.println("black bear");
    }
    // Run the spark program
    SparkManager sparkManager = applicationManager.getSparkManager(TransactionSpark.class.getSimpleName());
    sparkManager.start(ImmutableMap.of("input.file", inputFile.getAbsolutePath(), "keyvalue.table", "KeyValueTable", "result.all.dataset", "SparkResult", "result.threshold", "2", "result.threshold.dataset", "SparkThresholdResult"));
    // Verify result from dataset before the Spark program terminates
    final DataSetManager<KeyValueTable> resultManager = getDataset("SparkThresholdResult");
    final KeyValueTable resultTable = resultManager.get();
    // Expect the threshold result dataset, with threshold >=2, contains [brown, fox, bear]
    Tasks.waitFor(ImmutableSet.of("brown", "fox", "bear"), () -> {
        // This is to start a new TX
        resultManager.flush();
        LOG.info("Reading from threshold result");
        try (CloseableIterator<KeyValue<byte[], byte[]>> itor = resultTable.scan(null, null)) {
            return ImmutableSet.copyOf(Iterators.transform(itor, input -> {
                String word = Bytes.toString(input.getKey());
                LOG.info("{}, {}", word, Bytes.toInt(input.getValue()));
                return word;
            }));
        }
    }, 3, TimeUnit.MINUTES, 1, TimeUnit.SECONDS);
    sparkManager.stop();
    sparkManager.waitForRun(ProgramRunStatus.KILLED, 60, TimeUnit.SECONDS);
}