Examples with ETLConfig io.cdap.cdap.etl.proto.v2.ETLConfig used on opensource projects

Example 71 with ETLConfig

use of io.cdap.cdap.etl.proto.v2.ETLConfig in project cdap by caskdata.

the class DataStreamsSparkSinkTest method testSparkSink.

@Test
public // stream-rate-updater thread in Spark.
void testSparkSink() throws Exception {
    Schema schema = Schema.recordOf("test", Schema.Field.of("id", Schema.of(Schema.Type.STRING)), Schema.Field.of("name", Schema.of(Schema.Type.STRING)));
    List<StructuredRecord> input = new ArrayList<>();
    StructuredRecord samuelRecord = StructuredRecord.builder(schema).set("id", "0").set("name", "samuel").build();
    StructuredRecord jacksonRecord = StructuredRecord.builder(schema).set("id", "1").set("name", "jackson").build();
    StructuredRecord dwayneRecord = StructuredRecord.builder(schema).set("id", "2").set("name", "dwayne").build();
    StructuredRecord johnsonRecord = StructuredRecord.builder(schema).set("id", "3").set("name", "johnson").build();
    input.add(samuelRecord);
    input.add(jacksonRecord);
    input.add(dwayneRecord);
    input.add(johnsonRecord);
    DataStreamsConfig etlConfig = DataStreamsConfig.builder().addStage(new ETLStage("source", MockSource.getPlugin(schema, input))).addStage(new ETLStage("sink", io.cdap.cdap.etl.mock.spark.streaming.MockSink.getPlugin("${tablename}"))).addConnection("source", "sink").setCheckpointDir("file://" + TMP_FOLDER.getRoot().toPath().toString()).setBatchInterval("1s").build();
    ApplicationId appId = NamespaceId.DEFAULT.app("sparksinkapp");
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, etlConfig);
    ApplicationManager appManager = deployApplication(appId, appRequest);
    testSparkSink(appManager, "output1");
    testSparkSink(appManager, "output2");
}

Example 72 with ETLConfig

use of io.cdap.cdap.etl.proto.v2.ETLConfig in project cdap by caskdata.

the class PreviewDataStreamsTest method testDataStreamsPreviewRun.

@Test
public void testDataStreamsPreviewRun() throws Exception {
    PreviewManager previewManager = getPreviewManager();
    String sinkTableName = "singleOutput";
    Schema schema = Schema.recordOf("testRecord", Schema.Field.of("name", Schema.of(Schema.Type.STRING)));
    List<StructuredRecord> records = new ArrayList<>();
    StructuredRecord recordSamuel = StructuredRecord.builder(schema).set("name", "samuel").build();
    StructuredRecord recordBob = StructuredRecord.builder(schema).set("name", "bob").build();
    StructuredRecord recordTest = StructuredRecord.builder(schema).set("name", "test").build();
    records.add(recordSamuel);
    records.add(recordBob);
    records.add(recordTest);
    /*
     * source --> transform -> sink
     */
    DataStreamsConfig etlConfig = DataStreamsConfig.builder().addStage(new ETLStage("source", MockSource.getPlugin(schema, records))).addStage(new ETLStage("transform", IdentityTransform.getPlugin())).addStage(new ETLStage("sink", MockSink.getPlugin(sinkTableName))).addConnection("source", "transform").addConnection("transform", "sink").setNumOfRecordsPreview(100).setBatchInterval("1s").setCheckpointDir("file://" + TMP_FOLDER.getRoot().toPath().toString()).build();
    // Construct the preview config with the program name and program type.
    PreviewConfig previewConfig = new PreviewConfig(DataStreamsSparkLauncher.NAME, ProgramType.SPARK, Collections.<String, String>emptyMap(), 1);
    AppRequest<DataStreamsConfig> appRequest = new AppRequest<>(APP_ARTIFACT, etlConfig, previewConfig);
    // Start the preview and get the corresponding PreviewRunner.
    ApplicationId previewId = previewManager.start(NamespaceId.DEFAULT, appRequest);
    // Wait for the preview to be running and wait until the records are processed in the sink.
    Tasks.waitFor(true, new Callable<Boolean>() {

        @Override
        public Boolean call() throws Exception {
            Map<String, List<JsonElement>> data = previewManager.getData(previewId, "sink");
            return data != null && data.get(DATA_TRACER_PROPERTY) != null && data.get(DATA_TRACER_PROPERTY).size() == 3;
        }
    }, 1, TimeUnit.MINUTES);
    // check data in source and transform
    checkPreviewStore(previewManager, previewId, "source", 3);
    checkPreviewStore(previewManager, previewId, "transform", 3);
    // Wait for the pipeline to be shutdown by timer.
    TimeUnit.MINUTES.sleep(1);
    Tasks.waitFor(PreviewStatus.Status.KILLED_BY_TIMER, new Callable<PreviewStatus.Status>() {

        @Override
        public PreviewStatus.Status call() throws Exception {
            return previewManager.getStatus(previewId).getStatus();
        }
    }, 1, TimeUnit.MINUTES);
    // Validate the metrics for preview
    validateMetric(3, previewId, "source.records.out", previewManager);
    validateMetric(3, previewId, "transform.records.in", previewManager);
    validateMetric(3, previewId, "transform.records.out", previewManager);
    validateMetric(3, previewId, "sink.records.in", previewManager);
    validateMetric(3, previewId, "sink.records.out", previewManager);
    // Check the sink table is not created in the real space.
    DataSetManager<Table> sinkManager = getDataset(sinkTableName);
    Assert.assertNull(sinkManager.get());
}

Example 73 with ETLConfig

use of io.cdap.cdap.etl.proto.v2.ETLConfig in project cdap by caskdata.

the class PipelineSpecGenerator method validateConfig.

/**
 * Validate that this is a valid pipeline. A valid pipeline has the following properties:
 *
 * All stages in the pipeline have a unique name.
 * Source stages have at least one output and no inputs.
 * Sink stages have at least one input and no outputs.
 * There are no cycles in the pipeline.
 * All inputs into a stage have the same schema.
 * ErrorTransforms only have BatchSource, Transform, or BatchAggregator as input stages.
 * AlertPublishers have at least one input and no outputs and don't have SparkSink or BatchSink as input.
 * Action stages can only be at the start or end of the pipeline.
 * Condition stages have at most 2 outputs. Each stage on a condition's output branch has at most a single input.
 *
 * Returns the stages in the order they should be configured to ensure that all input stages are configured
 * before their output.
 *
 * @param config the user provided configuration
 * @return the order to configure the stages in
 * @throws IllegalArgumentException if the pipeline is invalid
 */
protected ValidatedPipeline validateConfig(ETLConfig config) {
    config.validate();
    if (config.getStages().isEmpty()) {
        throw new IllegalArgumentException("A pipeline must contain at least one stage.");
    }
    Set<String> actionStages = new HashSet<>();
    Set<String> conditionStages = new HashSet<>();
    Map<String, String> stageTypes = new HashMap<>();
    // check stage name uniqueness
    Set<String> stageNames = new HashSet<>();
    for (ETLStage stage : config.getStages()) {
        if (!stageNames.add(stage.getName())) {
            throw new IllegalArgumentException(String.format("Invalid pipeline. Multiple stages are named %s. Please ensure all stage names are unique", stage.getName()));
        }
        // if stage is Action stage, add it to the Action stage set
        if (isAction(stage.getPlugin().getType())) {
            actionStages.add(stage.getName());
        }
        // if the stage is condition add it to the Condition stage set
        if (stage.getPlugin().getType().equals(Condition.PLUGIN_TYPE)) {
            conditionStages.add(stage.getName());
        }
        stageTypes.put(stage.getName(), stage.getPlugin().getType());
    }
    // check that the from and to are names of actual stages
    // also check that conditions have at most 2 outgoing connections each label with true or
    // false but not both
    Map<String, Boolean> conditionBranch = new HashMap<>();
    for (Connection connection : config.getConnections()) {
        if (!stageNames.contains(connection.getFrom())) {
            throw new IllegalArgumentException(String.format("Invalid connection %s. %s is not a stage.", connection, connection.getFrom()));
        }
        if (!stageNames.contains(connection.getTo())) {
            throw new IllegalArgumentException(String.format("Invalid connection %s. %s is not a stage.", connection, connection.getTo()));
        }
        if (conditionStages.contains(connection.getFrom())) {
            if (connection.getCondition() == null) {
                String msg = String.format("For condition stage %s, the connection %s is not marked with either " + "'true' or 'false'.", connection.getFrom(), connection);
                throw new IllegalArgumentException(msg);
            }
            // check if connection from the condition node is marked as true or false multiple times
            if (conditionBranch.containsKey(connection.getFrom()) && connection.getCondition().equals(conditionBranch.get(connection.getFrom()))) {
                String msg = String.format("For condition stage '%s', more than one outgoing connections are marked as %s.", connection.getFrom(), connection.getCondition());
                throw new IllegalArgumentException(msg);
            }
            conditionBranch.put(connection.getFrom(), connection.getCondition());
        }
    }
    List<ETLStage> traversalOrder = new ArrayList<>(stageNames.size());
    // can only have empty connections if the pipeline consists of a single action.
    if (config.getConnections().isEmpty()) {
        if (actionStages.size() == 1 && stageNames.size() == 1) {
            traversalOrder.add(config.getStages().iterator().next());
            return new ValidatedPipeline(traversalOrder, config);
        } else {
            throw new IllegalArgumentException("Invalid pipeline. There are no connections between stages. " + "This is only allowed if the pipeline consists of a single action plugin.");
        }
    }
    Dag dag = new Dag(config.getConnections());
    Set<String> controlStages = Sets.union(actionStages, conditionStages);
    Map<String, ETLStage> stages = new HashMap<>();
    for (ETLStage stage : config.getStages()) {
        String stageName = stage.getName();
        Set<String> stageInputs = dag.getNodeInputs(stageName);
        Set<String> stageOutputs = dag.getNodeOutputs(stageName);
        String stageType = stage.getPlugin().getType();
        boolean isSource = isSource(stageType);
        boolean isSink = isSink(stageType);
        // check source plugins are sources in the dag
        if (isSource) {
            if (!stageInputs.isEmpty() && !controlStages.containsAll(stageInputs)) {
                throw new IllegalArgumentException(String.format("%s %s has incoming connections from %s. %s stages cannot have any incoming connections.", stageType, stageName, Joiner.on(',').join(stageInputs), stageType));
            }
            // check that source plugins are not present after any non-condition/action stage
            Set<String> parents = dag.parentsOf(stageName);
            Set<String> nonControlParents = Sets.difference(parents, controlStages);
            if (nonControlParents.size() > 1) {
                // the stage's nonControlParents should only contain itself
                throw new IllegalArgumentException(String.format("%s %s is invalid. %s stages can only be placed at the start of the pipeline.", stageType, stageName, stageType));
            }
        } else if (isSink) {
            if (!stageOutputs.isEmpty() && !controlStages.containsAll(stageOutputs)) {
                throw new IllegalArgumentException(String.format("%s %s has outgoing connections to %s. %s stages cannot have any outgoing connections.", stageType, stageName, Joiner.on(',').join(stageOutputs), stageType));
            }
        } else if (ErrorTransform.PLUGIN_TYPE.equals(stageType)) {
            for (String inputStage : stageInputs) {
                String inputType = stageTypes.get(inputStage);
                if (!VALID_ERROR_INPUTS.contains(inputType)) {
                    throw new IllegalArgumentException(String.format("ErrorTransform %s cannot have stage %s of type %s as input. Only %s stages can emit errors.", stageName, inputStage, inputType, Joiner.on(',').join(VALID_ERROR_INPUTS)));
                }
            }
        }
        boolean isAction = isAction(stageType);
        if (!isAction && !stageType.equals(Condition.PLUGIN_TYPE) && !isSource && stageInputs.isEmpty()) {
            throw new IllegalArgumentException(String.format("Stage %s is unreachable, it has no incoming connections.", stageName));
        }
        if (!isAction && !isSink && stageOutputs.isEmpty()) {
            throw new IllegalArgumentException(String.format("Stage %s is a dead end, it has no outgoing connections.", stageName));
        }
        stages.put(stageName, stage);
    }
    // make sure actions are not in the middle of the pipeline -- only at the start and/or end
    for (String actionStage : actionStages) {
        Set<String> actionParents = dag.parentsOf(actionStage);
        Set<String> actionChildren = dag.accessibleFrom(actionStage);
        Set<String> nonControlParents = Sets.difference(actionParents, controlStages);
        Set<String> nonControlChildren = Sets.difference(actionChildren, controlStages);
        if (!nonControlChildren.isEmpty() && !nonControlParents.isEmpty()) {
            throw new IllegalArgumentException(String.format("Action stage '%s' is invalid. Actions can only be placed at the start or end of the pipeline.", actionStage));
        }
    }
    validateConditionBranches(conditionStages, dag);
    for (String stageName : dag.getTopologicalOrder()) {
        traversalOrder.add(stages.get(stageName));
    }
    return new ValidatedPipeline(traversalOrder, config);
}

Example 74 with ETLConfig

use of io.cdap.cdap.etl.proto.v2.ETLConfig in project cdap by caskdata.

the class PipelineSpecGeneratorTest method testGenerateSpec.

@Test
public void testGenerateSpec() throws ValidationException {
    /*
     *           ---- t1 ------------
     *           |            |      |
     * source ---             |      |--- t3 --- sink1
     *           |            |      |
     *           ------------ t2 --------------- sink2
     *           |                        |
     *           |                        |
     *           -------------------------
     */
    ETLBatchConfig etlConfig = ETLBatchConfig.builder().setTimeSchedule("* * * * *").addStage(new ETLStage("source", MOCK_SOURCE)).addStage(new ETLStage("sink1", MOCK_SINK)).addStage(new ETLStage("sink2", MOCK_SINK)).addStage(new ETLStage("t1", MOCK_TRANSFORM_A)).addStage(new ETLStage("t2", MOCK_TRANSFORM_A)).addStage(new ETLStage("t3", MOCK_TRANSFORM_B)).addConnection("source", "t1").addConnection("source", "t2").addConnection("source", "sink2").addConnection("t1", "t2").addConnection("t1", "t3").addConnection("t1", "sink2").addConnection("t2", "sink2").addConnection("t2", "t3").addConnection("t3", "sink1").setNumOfRecordsPreview(100).build();
    // test the spec generated is correct, with the right input and output schemas and artifact information.
    BatchPipelineSpec actual = specGenerator.generateSpec(etlConfig);
    Map<String, String> emptyMap = ImmutableMap.of();
    PipelineSpec expected = BatchPipelineSpec.builder().addStage(StageSpec.builder("source", new PluginSpec(BatchSource.PLUGIN_TYPE, "mocksource", emptyMap, ARTIFACT_ID)).addOutput(SCHEMA_A, "t1", "t2", "sink2").build()).addStage(StageSpec.builder("sink1", new PluginSpec(BatchSink.PLUGIN_TYPE, "mocksink", emptyMap, ARTIFACT_ID)).addInputSchema("t3", SCHEMA_B).setErrorSchema(SCHEMA_B).build()).addStage(StageSpec.builder("sink2", new PluginSpec(BatchSink.PLUGIN_TYPE, "mocksink", emptyMap, ARTIFACT_ID)).addInputSchemas(ImmutableMap.of("t1", SCHEMA_A, "t2", SCHEMA_A, "source", SCHEMA_A)).setErrorSchema(SCHEMA_A).build()).addStage(StageSpec.builder("t1", new PluginSpec(Transform.PLUGIN_TYPE, "mockA", emptyMap, ARTIFACT_ID)).addInputSchema("source", SCHEMA_A).addOutput(SCHEMA_A, "t2", "t3", "sink2").setErrorSchema(SCHEMA_B).build()).addStage(StageSpec.builder("t2", new PluginSpec(Transform.PLUGIN_TYPE, "mockA", emptyMap, ARTIFACT_ID)).addInputSchemas(ImmutableMap.of("source", SCHEMA_A, "t1", SCHEMA_A)).addOutput(SCHEMA_A, "t3", "sink2").setErrorSchema(SCHEMA_B).build()).addStage(StageSpec.builder("t3", new PluginSpec(Transform.PLUGIN_TYPE, "mockB", emptyMap, ARTIFACT_ID)).addInputSchemas(ImmutableMap.of("t1", SCHEMA_A, "t2", SCHEMA_A)).addOutput(SCHEMA_B, "sink1").setErrorSchema(SCHEMA_A).build()).addConnections(etlConfig.getConnections()).setResources(etlConfig.getResources()).setDriverResources(new Resources(1024, 1)).setClientResources(new Resources(1024, 1)).setStageLoggingEnabled(etlConfig.isStageLoggingEnabled()).setNumOfRecordsPreview(etlConfig.getNumOfRecordsPreview()).build();
    Assert.assertEquals(expected, actual);
}

Example 75 with ETLConfig

use of io.cdap.cdap.etl.proto.v2.ETLConfig in project cdap by caskdata.

the class PipelineSpecGeneratorTest method testConflictingInputSchemasCondition.

@Test(expected = IllegalArgumentException.class)
public void testConflictingInputSchemasCondition() throws ValidationException {
    /*
     *           ---- transformA ----
     *           |                  |
     * source ---                   |--- condition -- sink
     *           |                  |
     *           ---- transformB ----
     *
     * sink gets schema A and schema B as input, should fail
     */
    ETLBatchConfig etlConfig = ETLBatchConfig.builder().setTimeSchedule("* * * * *").addStage(new ETLStage("source", MOCK_SOURCE)).addStage(new ETLStage("sink", MOCK_SINK)).addStage(new ETLStage("tA", MOCK_TRANSFORM_A)).addStage(new ETLStage("tB", MOCK_TRANSFORM_B)).addStage(new ETLStage("cond", MOCK_CONDITION)).addConnection("source", "tA").addConnection("source", "tB").addConnection("tA", "cond").addConnection("tB", "cond").addConnection("cond", "sink", true).build();
    specGenerator.generateSpec(etlConfig);
}