Examples with ProgramRunId co.cask.cdap.proto.id.ProgramRunId used on opensource projects

Example 26 with ProgramRunId

use of co.cask.cdap.proto.id.ProgramRunId in project cdap by caskdata.

the class SparkRuntimeContextProvider method createIfNotExists.

/**
   * Creates a singleton {@link SparkRuntimeContext}.
   * It has assumption on file location that are localized by the SparkRuntimeService.
   */
private static synchronized SparkRuntimeContext createIfNotExists() {
    if (sparkRuntimeContext != null) {
        return sparkRuntimeContext;
    }
    try {
        CConfiguration cConf = createCConf();
        Configuration hConf = createHConf();
        SparkRuntimeContextConfig contextConfig = new SparkRuntimeContextConfig(hConf);
        // Should be yarn only and only for executor node, not the driver node.
        Preconditions.checkState(!contextConfig.isLocal() && Boolean.parseBoolean(System.getenv("SPARK_YARN_MODE")), "SparkContextProvider.getSparkContext should only be called in Spark executor process.");
        // Create the program
        Program program = createProgram(cConf, contextConfig);
        Injector injector = createInjector(cConf, hConf, contextConfig.getProgramId(), contextConfig.getProgramOptions());
        final Service logAppenderService = new LogAppenderService(injector.getInstance(LogAppenderInitializer.class), contextConfig.getProgramOptions());
        final ZKClientService zkClientService = injector.getInstance(ZKClientService.class);
        final KafkaClientService kafkaClientService = injector.getInstance(KafkaClientService.class);
        final MetricsCollectionService metricsCollectionService = injector.getInstance(MetricsCollectionService.class);
        final StreamCoordinatorClient streamCoordinatorClient = injector.getInstance(StreamCoordinatorClient.class);
        // Use the shutdown hook to shutdown services, since this class should only be loaded from System classloader
        // of the spark executor, hence there should be exactly one instance only.
        // The problem with not shutting down nicely is that some logs/metrics might be lost
        Services.chainStart(logAppenderService, zkClientService, kafkaClientService, metricsCollectionService, streamCoordinatorClient);
        Runtime.getRuntime().addShutdownHook(new Thread() {

            @Override
            public void run() {
                // The logger may already been shutdown. Use System.out/err instead
                System.out.println("Shutting SparkClassLoader services");
                Future<List<ListenableFuture<Service.State>>> future = Services.chainStop(logAppenderService, streamCoordinatorClient, metricsCollectionService, kafkaClientService, zkClientService);
                try {
                    List<ListenableFuture<Service.State>> futures = future.get(5, TimeUnit.SECONDS);
                    System.out.println("SparkClassLoader services shutdown completed: " + futures);
                } catch (Exception e) {
                    System.err.println("Exception when shutting down services");
                    e.printStackTrace(System.err);
                }
            }
        });
        // Constructor the DatasetFramework
        DatasetFramework datasetFramework = injector.getInstance(DatasetFramework.class);
        WorkflowProgramInfo workflowInfo = contextConfig.getWorkflowProgramInfo();
        DatasetFramework programDatasetFramework = workflowInfo == null ? datasetFramework : NameMappedDatasetFramework.createFromWorkflowProgramInfo(datasetFramework, workflowInfo, contextConfig.getApplicationSpecification());
        // Setup dataset framework context, if required
        if (programDatasetFramework instanceof ProgramContextAware) {
            ProgramRunId programRunId = program.getId().run(ProgramRunners.getRunId(contextConfig.getProgramOptions()));
            ((ProgramContextAware) programDatasetFramework).setContext(new BasicProgramContext(programRunId));
        }
        PluginInstantiator pluginInstantiator = createPluginInstantiator(cConf, contextConfig, program.getClassLoader());
        // Create the context object
        sparkRuntimeContext = new SparkRuntimeContext(contextConfig.getConfiguration(), program, contextConfig.getProgramOptions(), cConf, getHostname(), injector.getInstance(TransactionSystemClient.class), programDatasetFramework, injector.getInstance(DiscoveryServiceClient.class), metricsCollectionService, injector.getInstance(StreamAdmin.class), contextConfig.getWorkflowProgramInfo(), pluginInstantiator, injector.getInstance(SecureStore.class), injector.getInstance(SecureStoreManager.class), injector.getInstance(AuthorizationEnforcer.class), injector.getInstance(AuthenticationContext.class), injector.getInstance(MessagingService.class));
        LoggingContextAccessor.setLoggingContext(sparkRuntimeContext.getLoggingContext());
        return sparkRuntimeContext;
    } catch (Exception e) {
        throw Throwables.propagate(e);
    }
}

Example 27 with ProgramRunId

use of co.cask.cdap.proto.id.ProgramRunId in project cdap by caskdata.

the class SparkExecutionServiceTest method testCompletion.

@Test
public void testCompletion() throws Exception {
    ProgramRunId programRunId = new ProgramRunId("ns", "app", ProgramType.SPARK, "test", RunIds.generate().getId());
    // Start a service that no token is supported
    SparkExecutionService service = new SparkExecutionService(locationFactory, InetAddress.getLoopbackAddress().getCanonicalHostName(), programRunId, null);
    service.startAndWait();
    try {
        SparkExecutionClient client = new SparkExecutionClient(service.getBaseURI(), programRunId);
        // Heartbeats multiple times.
        for (int i = 0; i < 5; i++) {
            Assert.assertNull(client.heartbeat(null));
            TimeUnit.MILLISECONDS.sleep(50);
        }
        // Call complete to notify the service it has been stopped
        client.completed(null);
    } finally {
        service.stopAndWait();
    }
}

Example 28 with ProgramRunId

use of co.cask.cdap.proto.id.ProgramRunId in project cdap by caskdata.

the class WorkflowHttpHandler method getWorkflowNodeStates.

@GET
@Path("/apps/{app-id}/workflows/{workflow-id}/runs/{run-id}/nodes/state")
public void getWorkflowNodeStates(HttpRequest request, HttpResponder responder, @PathParam("namespace-id") String namespaceId, @PathParam("app-id") String applicationId, @PathParam("workflow-id") String workflowId, @PathParam("run-id") String runId) throws NotFoundException {
    ApplicationId appId = Ids.namespace(namespaceId).app(applicationId);
    ApplicationSpecification appSpec = store.getApplication(appId);
    if (appSpec == null) {
        throw new ApplicationNotFoundException(appId);
    }
    ProgramId workflowProgramId = appId.workflow(workflowId);
    WorkflowSpecification workflowSpec = appSpec.getWorkflows().get(workflowProgramId.getProgram());
    if (workflowSpec == null) {
        throw new ProgramNotFoundException(workflowProgramId);
    }
    ProgramRunId workflowRunId = workflowProgramId.run(runId);
    if (store.getRun(workflowProgramId, runId) == null) {
        throw new NotFoundException(workflowRunId);
    }
    List<WorkflowNodeStateDetail> nodeStateDetails = store.getWorkflowNodeStates(workflowRunId);
    Map<String, WorkflowNodeStateDetail> nodeStates = new HashMap<>();
    for (WorkflowNodeStateDetail nodeStateDetail : nodeStateDetails) {
        nodeStates.put(nodeStateDetail.getNodeId(), nodeStateDetail);
    }
    responder.sendJson(HttpResponseStatus.OK, nodeStates, STRING_TO_NODESTATEDETAIL_MAP_TYPE);
}

Example 29 with ProgramRunId

use of co.cask.cdap.proto.id.ProgramRunId in project cdap by caskdata.

the class FlowletProgramRunner method run.

@SuppressWarnings("unchecked")
@Override
public ProgramController run(Program program, ProgramOptions options) {
    BasicFlowletContext flowletContext = null;
    try {
        // Extract and verify parameters
        String flowletName = options.getName();
        int instanceId = Integer.parseInt(options.getArguments().getOption(ProgramOptionConstants.INSTANCE_ID, "-1"));
        Preconditions.checkArgument(instanceId >= 0, "Missing instance Id");
        int instanceCount = Integer.parseInt(options.getArguments().getOption(ProgramOptionConstants.INSTANCES, "0"));
        Preconditions.checkArgument(instanceCount > 0, "Invalid or missing instance count");
        RunId runId = ProgramRunners.getRunId(options);
        ApplicationSpecification appSpec = program.getApplicationSpecification();
        Preconditions.checkNotNull(appSpec, "Missing application specification.");
        ProgramType processorType = program.getType();
        Preconditions.checkNotNull(processorType, "Missing processor type.");
        Preconditions.checkArgument(processorType == ProgramType.FLOW, "Only FLOW process type is supported.");
        String processorName = program.getName();
        Preconditions.checkNotNull(processorName, "Missing processor name.");
        FlowSpecification flowSpec = appSpec.getFlows().get(processorName);
        FlowletDefinition flowletDef = flowSpec.getFlowlets().get(flowletName);
        Preconditions.checkNotNull(flowletDef, "Definition missing for flowlet \"%s\"", flowletName);
        Class<?> clz = Class.forName(flowletDef.getFlowletSpec().getClassName(), true, program.getClassLoader());
        Preconditions.checkArgument(Flowlet.class.isAssignableFrom(clz), "%s is not a Flowlet.", clz);
        // Setup dataset framework context, if required
        ProgramId programId = program.getId();
        FlowletId flowletId = programId.flowlet(flowletName);
        ProgramRunId run = programId.run(runId);
        ProgramContext programContext = new BasicProgramContext(run, flowletId);
        if (dsFramework instanceof ProgramContextAware) {
            ((ProgramContextAware) dsFramework).setContext(programContext);
        }
        Class<? extends Flowlet> flowletClass = (Class<? extends Flowlet>) clz;
        // Creates flowlet context
        flowletContext = new BasicFlowletContext(program, options, flowletId, instanceId, instanceCount, flowletDef.getDatasets(), flowletDef.getFlowletSpec(), metricsCollectionService, discoveryServiceClient, txClient, dsFramework, secureStore, secureStoreManager, messageService, cConf);
        // Creates tx related objects
        DataFabricFacade dataFabricFacade = dataFabricFacadeFactory.create(program, flowletContext.getDatasetCache());
        if (dataFabricFacade instanceof ProgramContextAware) {
            ((ProgramContextAware) dataFabricFacade).setContext(programContext);
        }
        // Creates QueueSpecification
        Table<Node, String, Set<QueueSpecification>> queueSpecs = new SimpleQueueSpecificationGenerator(new ApplicationId(program.getNamespaceId(), program.getApplicationId())).create(flowSpec);
        Flowlet flowlet = new InstantiatorFactory(false).get(TypeToken.of(flowletClass)).create();
        TypeToken<? extends Flowlet> flowletType = TypeToken.of(flowletClass);
        // Set the context classloader to the cdap classloader. It is needed for the DatumWriterFactory be able
        // to load cdap classes
        Thread.currentThread().setContextClassLoader(FlowletProgramRunner.class.getClassLoader());
        // Inject DataSet, OutputEmitter, Metric fields
        ImmutableList.Builder<ProducerSupplier> queueProducerSupplierBuilder = ImmutableList.builder();
        Reflections.visit(flowlet, flowlet.getClass(), new PropertyFieldSetter(flowletDef.getFlowletSpec().getProperties()), new DataSetFieldSetter(flowletContext), new MetricsFieldSetter(flowletContext.getMetrics()), new OutputEmitterFieldSetter(outputEmitterFactory(flowletContext, flowletName, dataFabricFacade, queueProducerSupplierBuilder, queueSpecs)));
        ImmutableList.Builder<ConsumerSupplier<?>> queueConsumerSupplierBuilder = ImmutableList.builder();
        Collection<ProcessSpecification<?>> processSpecs = createProcessSpecification(flowletContext, flowletType, processMethodFactory(flowlet), processSpecificationFactory(flowletContext, dataFabricFacade, queueReaderFactory, flowletName, queueSpecs, queueConsumerSupplierBuilder, createSchemaCache(program)), Lists.<ProcessSpecification<?>>newLinkedList());
        List<ConsumerSupplier<?>> consumerSuppliers = queueConsumerSupplierBuilder.build();
        // Create the flowlet driver
        AtomicReference<FlowletProgramController> controllerRef = new AtomicReference<>();
        Service serviceHook = createServiceHook(flowletName, consumerSuppliers, controllerRef);
        FlowletRuntimeService driver = new FlowletRuntimeService(flowlet, flowletContext, processSpecs, createCallback(flowlet, flowletDef.getFlowletSpec()), dataFabricFacade, serviceHook);
        FlowletProgramController controller = new FlowletProgramController(program.getId(), flowletName, flowletContext, driver, queueProducerSupplierBuilder.build(), consumerSuppliers);
        controllerRef.set(controller);
        LOG.info("Starting flowlet: {}", flowletContext);
        driver.start();
        LOG.info("Flowlet started: {}", flowletContext);
        return controller;
    } catch (Exception e) {
        // of the flowlet context.
        if (flowletContext != null) {
            flowletContext.close();
        }
        throw Throwables.propagate(e);
    }
}

Example 30 with ProgramRunId

use of co.cask.cdap.proto.id.ProgramRunId in project cdap by caskdata.

the class BasicLineageWriterTest method testWrites.

@Test
public void testWrites() throws Exception {
    Injector injector = getInjector();
    MetadataStore metadataStore = injector.getInstance(MetadataStore.class);
    LineageStore lineageStore = injector.getInstance(LineageStore.class);
    LineageWriter lineageWriter = new BasicLineageWriter(lineageStore);
    // Define entities
    ProgramId program = new ProgramId(NamespaceId.DEFAULT.getNamespace(), "app", ProgramType.FLOW, "flow");
    StreamId stream = new StreamId(NamespaceId.DEFAULT.getNamespace(), "stream");
    ProgramRunId run1 = new ProgramRunId(program.getNamespace(), program.getApplication(), program.getType(), program.getEntityName(), RunIds.generate(10000).getId());
    ProgramRunId run2 = new ProgramRunId(program.getNamespace(), program.getApplication(), program.getType(), program.getEntityName(), RunIds.generate(20000).getId());
    // Tag stream
    metadataStore.addTags(MetadataScope.USER, stream, "stag1", "stag2");
    // Write access for run1
    lineageWriter.addAccess(run1, stream, AccessType.READ);
    Assert.assertEquals(ImmutableSet.of(program, stream), lineageStore.getEntitiesForRun(run1));
    // Record time to verify duplicate writes.
    long beforeSecondTag = System.currentTimeMillis();
    // Wait for next millisecond, since access time is stored in milliseconds.
    TimeUnit.MILLISECONDS.sleep(1);
    // Add another tag to stream
    metadataStore.addTags(MetadataScope.USER, stream, "stag3");
    // Write access for run1 again
    lineageWriter.addAccess(run1, stream, AccessType.READ);
    // The write should be no-op, and access time for run1 should not be updated
    Assert.assertTrue(lineageStore.getAccessTimesForRun(run1).get(0) < beforeSecondTag);
    // However, you can write access for another run
    lineageWriter.addAccess(run2, stream, AccessType.READ);
    // Assert new access time is written
    Assert.assertTrue(lineageStore.getAccessTimesForRun(run2).get(0) >= beforeSecondTag);
}