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

Example 26 with SparkManager

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

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 27 with SparkManager

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

the class DataStreamsTest method testJoin.

@Test
public void testJoin() throws Exception {
    /*
     * source1 ----> t1 ------
     *                        | --> innerjoin ----> t4 ------
     * source2 ----> t2 ------                                 |
     *                                                         | ---> outerjoin --> sink1
     *                                                         |
     * source3 -------------------- t3 ------------------------
     */
    Schema inputSchema1 = Schema.recordOf("customerRecord", Schema.Field.of("customer_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("customer_name", Schema.of(Schema.Type.STRING)));
    Schema inputSchema2 = Schema.recordOf("itemRecord", Schema.Field.of("item_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("item_price", Schema.of(Schema.Type.LONG)), Schema.Field.of("cust_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("cust_name", Schema.of(Schema.Type.STRING)));
    Schema inputSchema3 = Schema.recordOf("transactionRecord", Schema.Field.of("t_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("c_id", Schema.of(Schema.Type.STRING)), Schema.Field.of("i_id", Schema.of(Schema.Type.STRING)));
    Schema outSchema2 = Schema.recordOf("join.output", Schema.Field.of("t_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("c_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("i_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), 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.Field.of("item_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("item_price", Schema.nullableOf(Schema.of(Schema.Type.LONG))), Schema.Field.of("cust_id", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("cust_name", Schema.nullableOf(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 recordCar = StructuredRecord.builder(inputSchema2).set("item_id", "11").set("item_price", 10000L).set("cust_id", "1").set("cust_name", "samuel").build();
    StructuredRecord recordBike = StructuredRecord.builder(inputSchema2).set("item_id", "22").set("item_price", 100L).set("cust_id", "3").set("cust_name", "jane").build();
    StructuredRecord recordTrasCar = StructuredRecord.builder(inputSchema3).set("t_id", "1").set("c_id", "1").set("i_id", "11").build();
    StructuredRecord recordTrasBike = StructuredRecord.builder(inputSchema3).set("t_id", "2").set("c_id", "3").set("i_id", "22").build();
    StructuredRecord recordTrasPlane = StructuredRecord.builder(inputSchema3).set("t_id", "3").set("c_id", "4").set("i_id", "33").build();
    List<StructuredRecord> input1 = ImmutableList.of(recordSamuel, recordBob, recordJane);
    List<StructuredRecord> input2 = ImmutableList.of(recordCar, recordBike);
    List<StructuredRecord> input3 = ImmutableList.of(recordTrasCar, recordTrasBike, recordTrasPlane);
    String outputName = "multiJoinOutputSink";
    DataStreamsConfig etlConfig = DataStreamsConfig.builder().addStage(new ETLStage("source1", MockSource.getPlugin(inputSchema1, input1))).addStage(new ETLStage("source2", MockSource.getPlugin(inputSchema2, input2))).addStage(new ETLStage("source3", MockSource.getPlugin(inputSchema3, input3))).addStage(new ETLStage("t1", IdentityTransform.getPlugin())).addStage(new ETLStage("t2", IdentityTransform.getPlugin())).addStage(new ETLStage("t3", IdentityTransform.getPlugin())).addStage(new ETLStage("t4", IdentityTransform.getPlugin())).addStage(new ETLStage("innerjoin", MockJoiner.getPlugin("t1.customer_id=t2.cust_id", "t1,t2", ""))).addStage(new ETLStage("outerjoin", MockJoiner.getPlugin("t4.item_id=t3.i_id", "", ""))).addStage(new ETLStage("multijoinSink", MockSink.getPlugin(outputName))).addConnection("source1", "t1").addConnection("source2", "t2").addConnection("source3", "t3").addConnection("t1", "innerjoin").addConnection("t2", "innerjoin").addConnection("innerjoin", "t4").addConnection("t3", "outerjoin").addConnection("t4", "outerjoin").addConnection("outerjoin", "multijoinSink").setBatchInterval("5s").setCheckpointDir(checkpointDir).build();
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, etlConfig);
    ApplicationId appId = NamespaceId.DEFAULT.app("JoinerApp");
    ApplicationManager appManager = deployApplication(appId, appRequest);
    SparkManager sparkManager = appManager.getSparkManager(DataStreamsSparkLauncher.NAME);
    sparkManager.start();
    sparkManager.waitForRun(ProgramRunStatus.RUNNING, 10, TimeUnit.SECONDS);
    StructuredRecord joinRecordSamuel = StructuredRecord.builder(outSchema2).set("customer_id", "1").set("customer_name", "samuel").set("item_id", "11").set("item_price", 10000L).set("cust_id", "1").set("cust_name", "samuel").set("t_id", "1").set("c_id", "1").set("i_id", "11").build();
    StructuredRecord joinRecordJane = StructuredRecord.builder(outSchema2).set("customer_id", "3").set("customer_name", "jane").set("item_id", "22").set("item_price", 100L).set("cust_id", "3").set("cust_name", "jane").set("t_id", "2").set("c_id", "3").set("i_id", "22").build();
    StructuredRecord joinRecordPlane = StructuredRecord.builder(outSchema2).set("t_id", "3").set("c_id", "4").set("i_id", "33").build();
    Set<StructuredRecord> expected = ImmutableSet.of(joinRecordSamuel, joinRecordJane, joinRecordPlane);
    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);
    validateMetric(appId, "source1.records.out", 3);
    validateMetric(appId, "source2.records.out", 2);
    validateMetric(appId, "source3.records.out", 3);
    validateMetric(appId, "t1.records.in", 3);
    validateMetric(appId, "t1.records.out", 3);
    validateMetric(appId, "t2.records.in", 2);
    validateMetric(appId, "t2.records.out", 2);
    validateMetric(appId, "t3.records.in", 3);
    validateMetric(appId, "t3.records.out", 3);
    validateMetric(appId, "t4.records.in", 2);
    validateMetric(appId, "t4.records.out", 2);
    validateMetric(appId, "innerjoin.records.in", 5);
    validateMetric(appId, "innerjoin.records.out", 2);
    validateMetric(appId, "outerjoin.records.in", 5);
    validateMetric(appId, "outerjoin.records.out", 3);
    validateMetric(appId, "multijoinSink.records.in", 3);
}

Example 28 with SparkManager

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

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);
}

Example 29 with SparkManager

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

the class DataStreamsTest method testLineageWithMacro.

private RunId testLineageWithMacro(ApplicationManager appManager, Set<StructuredRecord> expected, String outputName) throws Exception {
    SparkManager sparkManager = appManager.getSparkManager(DataStreamsSparkLauncher.NAME);
    sparkManager.start(Collections.singletonMap("output", outputName));
    sparkManager.waitForRun(ProgramRunStatus.RUNNING, 10, TimeUnit.SECONDS);
    // since dataset name is a macro, the dataset isn't created until it is needed. Wait for it to exist
    Tasks.waitFor(true, () -> getDataset(outputName).get() != null, 1, TimeUnit.MINUTES);
    DataSetManager<Table> outputManager = getDataset(outputName);
    Tasks.waitFor(true, () -> {
        outputManager.flush();
        Set<StructuredRecord> outputRecords = new HashSet<>(MockSink.readOutput(outputManager));
        return expected.equals(outputRecords);
    }, 1, TimeUnit.MINUTES);
    sparkManager.stop();
    sparkManager.waitForStopped(10, TimeUnit.SECONDS);
    return RunIds.fromString(sparkManager.getHistory().iterator().next().getPid());
}

Example 30 with SparkManager

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

the class DataStreamsTest method testSplitterTransform.

@Test
public void testSplitterTransform() throws Exception {
    Schema schema = Schema.recordOf("user", Schema.Field.of("id", Schema.of(Schema.Type.LONG)), Schema.Field.of("name", Schema.nullableOf(Schema.of(Schema.Type.STRING))), Schema.Field.of("email", Schema.nullableOf(Schema.of(Schema.Type.STRING))));
    StructuredRecord user0 = StructuredRecord.builder(schema).set("id", 0L).build();
    StructuredRecord user1 = StructuredRecord.builder(schema).set("id", 1L).set("email", "one@example.com").build();
    StructuredRecord user2 = StructuredRecord.builder(schema).set("id", 2L).set("name", "two").build();
    StructuredRecord user3 = StructuredRecord.builder(schema).set("id", 3L).set("name", "three").set("email", "three@example.com").build();
    File outputDir = TMP_FOLDER.newFolder();
    String output1 = new File(outputDir, "output1").getAbsolutePath();
    String output2 = new File(outputDir, "output2").getAbsolutePath();
    /*
     *
     *                                            |null --> sink1
     *                       |null--> splitter2 --|
     * source --> splitter1--|                    |non-null --|
     *                       |                                |--> sink2
     *                       |non-null------------------------|
     */
    DataStreamsConfig config = DataStreamsConfig.builder().setBatchInterval("5s").addStage(new ETLStage("source", MockSource.getPlugin(schema, ImmutableList.of(user0, user1, user2, user3)))).addStage(new ETLStage("splitter1", NullFieldSplitterTransform.getPlugin("name"))).addStage(new ETLStage("splitter2", NullFieldSplitterTransform.getPlugin("email"))).addStage(new ETLStage("sink1", MockExternalSink.getPlugin(UUID.randomUUID().toString(), "sink1", output1))).addStage(new ETLStage("sink2", MockExternalSink.getPlugin(UUID.randomUUID().toString(), "sink2", output2))).addConnection("source", "splitter1").addConnection("splitter1", "splitter2", "null").addConnection("splitter1", "sink2", "non-null").addConnection("splitter2", "sink1", "null").addConnection("splitter2", "sink2", "non-null").setCheckpointDir(checkpointDir).build();
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, config);
    ApplicationId appId = NamespaceId.DEFAULT.app("SplitterTest");
    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, 10, TimeUnit.SECONDS);
    // check output
    // sink1 should only have records where both name and email are null (user0)
    Set<StructuredRecord> expected1 = ImmutableSet.of(user0);
    Tasks.waitFor(true, () -> {
        Set<StructuredRecord> outputRecords = new HashSet<>(MockExternalSink.readOutput(output1, schema));
        return expected1.equals(outputRecords);
    }, 4, TimeUnit.MINUTES);
    // sink2 should have anything with a non-null name or non-null email
    Set<StructuredRecord> expected2 = ImmutableSet.of(user1, user2, user3);
    Tasks.waitFor(true, () -> {
        Set<StructuredRecord> outputRecords = new HashSet<>(MockExternalSink.readOutput(output2, schema));
        return expected2.equals(outputRecords);
    }, 4, TimeUnit.MINUTES);
    sparkManager.stop();
    sparkManager.waitForStopped(10, TimeUnit.SECONDS);
    validateMetric(appId, "source.records.out", 4);
    validateMetric(appId, "splitter1.records.in", 4);
    validateMetric(appId, "splitter1.records.out.non-null", 2);
    validateMetric(appId, "splitter1.records.out.null", 2);
    validateMetric(appId, "splitter2.records.in", 2);
    validateMetric(appId, "splitter2.records.out.non-null", 1);
    validateMetric(appId, "splitter2.records.out.null", 1);
    validateMetric(appId, "sink1.records.in", 1);
    validateMetric(appId, "sink2.records.in", 3);
}