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

Example 21 with SparkManager

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

the class DataStreamsTest method testAlertPublisher.

@Test
public void testAlertPublisher() throws Exception {
    String sinkName = "alertSink";
    String topic = "alertTopic";
    Schema schema = Schema.recordOf("x", Schema.Field.of("id", Schema.nullableOf(Schema.of(Schema.Type.LONG))));
    StructuredRecord record1 = StructuredRecord.builder(schema).set("id", 1L).build();
    StructuredRecord record2 = StructuredRecord.builder(schema).set("id", 2L).build();
    StructuredRecord alertRecord = StructuredRecord.builder(schema).build();
    /*
     * source --> nullAlert --> sink
     *               |
     *               |--> TMS publisher
     */
    DataStreamsConfig config = DataStreamsConfig.builder().setBatchInterval("5s").addStage(new ETLStage("source", MockSource.getPlugin(schema, ImmutableList.of(record1, record2, alertRecord)))).addStage(new ETLStage("nullAlert", NullAlertTransform.getPlugin("id"))).addStage(new ETLStage("sink", MockSink.getPlugin(sinkName))).addStage(new ETLStage("tms", TMSAlertPublisher.getPlugin(topic, NamespaceId.DEFAULT.getNamespace()))).addConnection("source", "nullAlert").addConnection("nullAlert", "sink").addConnection("nullAlert", "tms").setCheckpointDir(checkpointDir).build();
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, config);
    ApplicationId appId = NamespaceId.DEFAULT.app("AlertTest");
    ApplicationManager appManager = deployApplication(appId, appRequest);
    SparkManager sparkManager = appManager.getSparkManager(DataStreamsSparkLauncher.NAME);
    sparkManager.start();
    sparkManager.waitForRun(ProgramRunStatus.RUNNING, 10, TimeUnit.SECONDS);
    final Set<StructuredRecord> expectedRecords = ImmutableSet.of(record1, record2);
    final Set<Alert> expectedMessages = ImmutableSet.of(new Alert("nullAlert", new HashMap<String, String>()));
    final DataSetManager<Table> sinkTable = getDataset(sinkName);
    Tasks.waitFor(true, () -> {
        // get alerts from TMS
        try {
            getMessagingAdmin(NamespaceId.DEFAULT.getNamespace()).getTopicProperties(topic);
        } catch (TopicNotFoundException e) {
            return false;
        }
        MessageFetcher messageFetcher = getMessagingContext().getMessageFetcher();
        Set<Alert> actualMessages = new HashSet<>();
        try (CloseableIterator<Message> iter = messageFetcher.fetch(NamespaceId.DEFAULT.getNamespace(), topic, 5, 0)) {
            while (iter.hasNext()) {
                Message message = iter.next();
                Alert alert = message.decodePayload(r -> GSON.fromJson(r, Alert.class));
                actualMessages.add(alert);
            }
        }
        // get records from sink
        sinkTable.flush();
        Set<StructuredRecord> outputRecords = new HashSet<>(MockSink.readOutput(sinkTable));
        return expectedRecords.equals(outputRecords) && expectedMessages.equals(actualMessages);
    }, 4, TimeUnit.MINUTES);
    sparkManager.stop();
    sparkManager.waitForStopped(10, TimeUnit.SECONDS);
    validateMetric(appId, "source.records.out", 3);
    validateMetric(appId, "nullAlert.records.in", 3);
    validateMetric(appId, "nullAlert.records.out", 2);
    validateMetric(appId, "nullAlert.records.alert", 1);
    validateMetric(appId, "sink.records.in", 2);
    validateMetric(appId, "tms.records.in", 1);
}

Example 22 with SparkManager

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

the class DataStreamsTest method testAutoJoinNullEquality.

private void testAutoJoinNullEquality(boolean nullSafe) throws Exception {
    /*
     * customers ----------|
     *                     |
     *                     |---> join ---> sink
     *                     |
     * transactions -------|
     */
    Schema inputSchema1 = Schema.recordOf("customer", Schema.Field.of("customer_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("customer_name", Schema.nullableOf(Schema.of(Schema.Type.STRING))));
    Schema inputSchema2 = Schema.recordOf("transaction", Schema.Field.of("t_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("customer_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("item_id", Schema.of(Schema.Type.STRING)));
    Schema outSchema = Schema.recordOf("customers.transactions", Schema.Field.of("customers_customer_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("customers_customer_name", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("transactions_t_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("transactions_customer_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("transactions_item_id", Schema.of(Schema.Type.STRING)));
    StructuredRecord recordSamuel = StructuredRecord.builder(inputSchema1).set("customer_id", "1").set("customer_name", "samuel").build();
    StructuredRecord recordBob = StructuredRecord.builder(inputSchema1).set("customer_name", "bob").build();
    StructuredRecord recordJane = StructuredRecord.builder(inputSchema1).set("customer_id", "3").set("customer_name", "jane").build();
    StructuredRecord trans1 = StructuredRecord.builder(inputSchema2).set("t_id", "1").set("customer_id", "1").set("item_id", "11").build();
    StructuredRecord trans2 = StructuredRecord.builder(inputSchema2).set("t_id", "2").set("customer_id", "3").set("item_id", "22").build();
    StructuredRecord trans3 = StructuredRecord.builder(inputSchema2).set("t_id", "3").set("item_id", "33").build();
    List<StructuredRecord> input1 = ImmutableList.of(recordSamuel, recordBob, recordJane);
    List<StructuredRecord> input2 = ImmutableList.of(trans1, trans2, trans3);
    String outputName = UUID.randomUUID().toString();
    DataStreamsConfig etlConfig = DataStreamsConfig.builder().addStage(new ETLStage("customers", MockSource.getPlugin(inputSchema1, input1))).addStage(new ETLStage("transactions", MockSource.getPlugin(inputSchema2, input2))).addStage(new ETLStage("join", MockAutoJoiner.getPlugin(Arrays.asList("customers", "transactions"), Collections.singletonList("customer_id"), Collections.singletonList("transactions"), Collections.emptyList(), Collections.emptyList(), nullSafe))).addStage(new ETLStage("sink", MockSink.getPlugin(outputName))).addConnection("customers", "join").addConnection("transactions", "join").addConnection("join", "sink").setBatchInterval("5s").setCheckpointDir(checkpointDir).build();
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, etlConfig);
    ApplicationId appId = NamespaceId.DEFAULT.app(UUID.randomUUID().toString());
    ApplicationManager appManager = deployApplication(appId, appRequest);
    SparkManager sparkManager = appManager.getSparkManager(DataStreamsSparkLauncher.NAME);
    sparkManager.start();
    sparkManager.waitForRun(ProgramRunStatus.RUNNING, 10, TimeUnit.SECONDS);
    StructuredRecord join1 = StructuredRecord.builder(outSchema).set("customers_customer_id", "1").set("customers_customer_name", "samuel").set("transactions_t_id", "1").set("transactions_customer_id", "1").set("transactions_item_id", "11").build();
    StructuredRecord join2 = StructuredRecord.builder(outSchema).set("customers_customer_id", "3").set("customers_customer_name", "jane").set("transactions_t_id", "2").set("transactions_customer_id", "3").set("transactions_item_id", "22").build();
    StructuredRecord join3;
    if (nullSafe) {
        // this transaction has a null customer id, which should match with the null id from customers
        join3 = StructuredRecord.builder(outSchema).set("transactions_t_id", "3").set("transactions_item_id", "33").set("customers_customer_name", "bob").build();
    } else {
        // this transaction has a null customer id, which should not match with the null id from customers
        join3 = StructuredRecord.builder(outSchema).set("transactions_t_id", "3").set("transactions_item_id", "33").build();
    }
    Set<StructuredRecord> expected = ImmutableSet.of(join1, join2, join3);
    DataSetManager<Table> outputManager = getDataset(outputName);
    Tasks.waitFor(true, () -> {
        outputManager.flush();
        Set<StructuredRecord> outputRecords = new HashSet<>(MockSink.readOutput(outputManager));
        return expected.equals(outputRecords);
    }, 4, TimeUnit.MINUTES);
    sparkManager.stop();
    sparkManager.waitForStopped(10, TimeUnit.SECONDS);
}

Example 23 with SparkManager

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

the class DataStreamsTest method testWindower.

@Test
public void testWindower() throws Exception {
    /*
     * source --> window(width=10,interval=1) --> aggregator --> filter --> sink
     */
    Schema schema = Schema.recordOf("data", Schema.Field.of("x", Schema.of(Schema.Type.STRING)));
    List<StructuredRecord> input = ImmutableList.of(StructuredRecord.builder(schema).set("x", "abc").build(), StructuredRecord.builder(schema).set("x", "abc").build(), StructuredRecord.builder(schema).set("x", "abc").build());
    String sinkName = "windowOut";
    // source sleeps 1 second between outputs
    DataStreamsConfig etlConfig = DataStreamsConfig.builder().addStage(new ETLStage("source", MockSource.getPlugin(schema, input, 1000L))).addStage(new ETLStage("window", Window.getPlugin(30, 1))).addStage(new ETLStage("agg", FieldCountAggregator.getPlugin("x", "string"))).addStage(new ETLStage("filter", StringValueFilterTransform.getPlugin("x", "all"))).addStage(new ETLStage("sink", MockSink.getPlugin(sinkName))).addConnection("source", "window").addConnection("window", "agg").addConnection("agg", "filter").addConnection("filter", "sink").setBatchInterval("1s").setCheckpointDir(checkpointDir).build();
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, etlConfig);
    ApplicationId appId = NamespaceId.DEFAULT.app("WindowerApp");
    ApplicationManager appManager = deployApplication(appId, appRequest);
    SparkManager sparkManager = appManager.getSparkManager(DataStreamsSparkLauncher.NAME);
    sparkManager.start();
    sparkManager.waitForRun(ProgramRunStatus.RUNNING, 10, TimeUnit.SECONDS);
    // the sink should contain at least one record with count of 3, and no records with more than 3.
    // less than 3 if the window doesn't contain all 3 records yet, but there should eventually be a window
    // that contains all 3.
    final DataSetManager<Table> outputManager = getDataset(sinkName);
    Tasks.waitFor(true, new Callable<Boolean>() {

        @Override
        public Boolean call() throws Exception {
            outputManager.flush();
            boolean sawThree = false;
            for (StructuredRecord record : MockSink.readOutput(outputManager)) {
                long count = record.get("ct");
                if (count == 3L) {
                    sawThree = true;
                }
                Assert.assertTrue(count <= 3L);
            }
            return sawThree;
        }
    }, 2, TimeUnit.MINUTES);
    sparkManager.stop();
}

Example 24 with SparkManager

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

the class DataStreamsTest method testAutoJoin.

@Test
public void testAutoJoin() throws Exception {
    /*
     * customers ----------|
     *                     |
     *                     |---> join ---> sink
     *                     |
     * transactions -------|
     */
    Schema inputSchema1 = Schema.recordOf("customer", Schema.Field.of("customer_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("customer_name", Schema.of(Schema.Type.STRING)));
    Schema inputSchema2 = Schema.recordOf("transaction", Schema.Field.of("t_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("customer_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("item_id", Schema.of(Schema.Type.STRING)));
    Schema outSchema = Schema.recordOf("customers.transactions", Schema.Field.of("customers_customer_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("customers_customer_name", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("transactions_t_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("transactions_customer_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("transactions_item_id", Schema.of(Schema.Type.STRING)));
    StructuredRecord recordSamuel = StructuredRecord.builder(inputSchema1).set("customer_id", "1").set("customer_name", "samuel").build();
    StructuredRecord recordBob = StructuredRecord.builder(inputSchema1).set("customer_id", "2").set("customer_name", "bob").build();
    StructuredRecord recordJane = StructuredRecord.builder(inputSchema1).set("customer_id", "3").set("customer_name", "jane").build();
    StructuredRecord tx1 = StructuredRecord.builder(inputSchema2).set("t_id", "1").set("customer_id", "1").set("item_id", "11").build();
    StructuredRecord tx2 = StructuredRecord.builder(inputSchema2).set("t_id", "2").set("customer_id", "3").set("item_id", "22").build();
    StructuredRecord tx3 = StructuredRecord.builder(inputSchema2).set("t_id", "3").set("customer_id", "4").set("item_id", "33").build();
    List<StructuredRecord> input1 = ImmutableList.of(recordSamuel, recordBob, recordJane);
    List<StructuredRecord> input2 = ImmutableList.of(tx1, tx2, tx3);
    String outputName = UUID.randomUUID().toString();
    DataStreamsConfig etlConfig = DataStreamsConfig.builder().addStage(new ETLStage("customers", MockSource.getPlugin(inputSchema1, input1))).addStage(new ETLStage("transactions", MockSource.getPlugin(inputSchema2, input2))).addStage(new ETLStage("join", MockAutoJoiner.getPlugin(Arrays.asList("customers", "transactions"), Collections.singletonList("customer_id"), Collections.singletonList("transactions"), Collections.emptyList(), Collections.emptyList(), true))).addStage(new ETLStage("sink", MockSink.getPlugin(outputName))).addConnection("customers", "join").addConnection("transactions", "join").addConnection("join", "sink").setBatchInterval("5s").setCheckpointDir(checkpointDir).build();
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, etlConfig);
    ApplicationId appId = NamespaceId.DEFAULT.app("AutoJoinerApp");
    ApplicationManager appManager = deployApplication(appId, appRequest);
    SparkManager sparkManager = appManager.getSparkManager(DataStreamsSparkLauncher.NAME);
    sparkManager.start();
    sparkManager.waitForRun(ProgramRunStatus.RUNNING, 10, TimeUnit.SECONDS);
    StructuredRecord join1 = StructuredRecord.builder(outSchema).set("customers_customer_id", "1").set("customers_customer_name", "samuel").set("transactions_t_id", "1").set("transactions_customer_id", "1").set("transactions_item_id", "11").build();
    StructuredRecord join2 = StructuredRecord.builder(outSchema).set("customers_customer_id", "3").set("customers_customer_name", "jane").set("transactions_t_id", "2").set("transactions_customer_id", "3").set("transactions_item_id", "22").build();
    StructuredRecord join3 = StructuredRecord.builder(outSchema).set("transactions_t_id", "3").set("transactions_customer_id", "4").set("transactions_item_id", "33").build();
    Set<StructuredRecord> expected = ImmutableSet.of(join1, join2, join3);
    DataSetManager<Table> outputManager = getDataset(outputName);
    Tasks.waitFor(true, () -> {
        outputManager.flush();
        Set<StructuredRecord> outputRecords = new HashSet<>(MockSink.readOutput(outputManager));
        return expected.equals(outputRecords);
    }, 4, TimeUnit.MINUTES);
    sparkManager.stop();
    sparkManager.waitForStopped(10, TimeUnit.SECONDS);
}

Example 25 with SparkManager

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

the class DataStreamsTest method testStageConsolidation.

@Test
public void testStageConsolidation() throws Exception {
    /*
                                   |non-null port --> sink1
        items1 --> null splitter --|
                                   |null port--> sink2
                                                   ^
                 |--> filter out id == 0  ---------|
        items2 --|    |
                 |    |--> error collector --> sink3
                 |
                 |                 |--> filter out id == 1 --> sink4
                 |--> aggregator --|
                                   |--> filter out id == 2 --> sink5
     */
    Schema schema = Schema.recordOf("item", Schema.Field.of("id", Schema.nullableOf(Schema.of(Schema.Type.INT))));
    StructuredRecord itemNull = StructuredRecord.builder(schema).build();
    StructuredRecord item0 = StructuredRecord.builder(schema).set("id", 0).build();
    StructuredRecord item1 = StructuredRecord.builder(schema).set("id", 1).build();
    StructuredRecord item2 = StructuredRecord.builder(schema).set("id", 2).build();
    StructuredRecord item3 = StructuredRecord.builder(schema).set("id", 3).build();
    List<StructuredRecord> input1Records = Arrays.asList(itemNull, item3);
    List<StructuredRecord> input2Records = Arrays.asList(item0, item1, item2);
    File outputDir = TMP_FOLDER.newFolder();
    String output1 = new File(outputDir, "output1").getAbsolutePath();
    String output2 = new File(outputDir, "output2").getAbsolutePath();
    String output3 = new File(outputDir, "output3").getAbsolutePath();
    String output4 = new File(outputDir, "output4").getAbsolutePath();
    String output5 = new File(outputDir, "output5").getAbsolutePath();
    DataStreamsConfig config = DataStreamsConfig.builder().addStage(new ETLStage("items1", MockSource.getPlugin(schema, input1Records))).addStage(new ETLStage("items2", MockSource.getPlugin(schema, input2Records))).addStage(new ETLStage("nullSplitter", NullFieldSplitterTransform.getPlugin("id"))).addStage(new ETLStage("filter0", IntValueFilterTransform.getPlugin("id", 0))).addStage(new ETLStage("collector", FlattenErrorTransform.getPlugin())).addStage(new ETLStage("filter1", IntValueFilterTransform.getPlugin("id", 1))).addStage(new ETLStage("filter2", IntValueFilterTransform.getPlugin("id", 2))).addStage(new ETLStage("identityAggregator", IdentityAggregator.getPlugin())).addStage(new ETLStage("sink1", MockExternalSink.getPlugin(UUID.randomUUID().toString(), "s1", output1))).addStage(new ETLStage("sink2", MockExternalSink.getPlugin(UUID.randomUUID().toString(), "s2", output2))).addStage(new ETLStage("sink3", MockExternalSink.getPlugin(UUID.randomUUID().toString(), "s3", output3))).addStage(new ETLStage("sink4", MockExternalSink.getPlugin(UUID.randomUUID().toString(), "s4", output4))).addStage(new ETLStage("sink5", MockExternalSink.getPlugin(UUID.randomUUID().toString(), "s5", output5))).addConnection("items1", "nullSplitter").addConnection("nullSplitter", "sink1", "non-null").addConnection("nullSplitter", "sink2", "null").addConnection("items2", "filter0").addConnection("items2", "identityAggregator").addConnection("filter0", "sink2").addConnection("filter0", "collector").addConnection("identityAggregator", "filter1").addConnection("identityAggregator", "filter2").addConnection("collector", "sink3").addConnection("filter1", "sink4").addConnection("filter2", "sink5").setProperties(Collections.singletonMap(io.cdap.cdap.etl.common.Constants.SPARK_PIPELINE_AUTOCACHE_ENABLE_FLAG, "false")).build();
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, config);
    ApplicationId appId = NamespaceId.DEFAULT.app("StageConsolidationTest");
    ApplicationManager appManager = deployApplication(appId, appRequest);
    // run pipeline
    Map<String, String> args = Collections.singletonMap(io.cdap.cdap.etl.common.Constants.CONSOLIDATE_STAGES, "true");
    SparkManager sparkManager = appManager.getSparkManager(DataStreamsSparkLauncher.NAME);
    sparkManager.startAndWaitForGoodRun(args, ProgramRunStatus.RUNNING, 5, TimeUnit.MINUTES);
    Schema errorSchema = Schema.recordOf("erroritem", Schema.Field.of("id", Schema.nullableOf(Schema.of(Schema.Type.INT))), Schema.Field.of("errMsg", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("errCode", Schema.nullableOf(Schema.of(Schema.Type.INT))), Schema.Field.of("errStage", Schema.nullableOf(Schema.of(Schema.Type.STRING))));
    StructuredRecord expectedError = StructuredRecord.builder(errorSchema).set("id", 0).set("errMsg", IntValueFilterTransform.ERROR_MESSAGE).set("errCode", IntValueFilterTransform.ERROR_CODE).set("errStage", "filter0").build();
    Set<StructuredRecord> sink1Expected = Collections.singleton(item3);
    Set<StructuredRecord> sink2Expected = new HashSet<>(Arrays.asList(itemNull, item1, item2));
    Set<StructuredRecord> sink3Expected = Collections.singleton(expectedError);
    Set<StructuredRecord> sink4Expected = new HashSet<>(Arrays.asList(item0, item2));
    Set<StructuredRecord> sink5Expected = new HashSet<>(Arrays.asList(item0, item1));
    Tasks.waitFor(true, () -> sink1Expected.equals(new HashSet<>(MockExternalSink.readOutput(output1, schema))) && sink2Expected.equals(new HashSet<>(MockExternalSink.readOutput(output2, schema))) && sink3Expected.equals(new HashSet<>(MockExternalSink.readOutput(output3, errorSchema))) && sink4Expected.equals(new HashSet<>(MockExternalSink.readOutput(output4, schema))) && sink5Expected.equals(new HashSet<>(MockExternalSink.readOutput(output5, schema))), 3, TimeUnit.MINUTES);
    sparkManager.stop();
    sparkManager.waitForStopped(1, TimeUnit.MINUTES);
    // check output
    validateMetric(appId, "nullSplitter.records.in", 2);
    validateMetric(appId, "nullSplitter.records.out.null", 1);
    validateMetric(appId, "nullSplitter.records.out.non-null", 1);
    validateMetric(appId, "filter0.records.out", 2);
    validateMetric(appId, "filter0.records.error", 1);
    validateMetric(appId, "identityAggregator.records.in", 3);
    validateMetric(appId, "identityAggregator.records.out", 3);
}