Search in sources :

Example 6 with SparkSpecification

use of io.cdap.cdap.api.spark.SparkSpecification in project cdap by cdapio.

the class SparkProgramRunner method run.

@Override
public ProgramController run(Program program, ProgramOptions options) {
    LOG.trace("Starting Spark program {} with SparkProgramRunner of ClassLoader {}", program.getId(), getClass().getClassLoader());
    // Get the RunId first. It is used for the creation of the ClassLoader closing thread.
    Arguments arguments = options.getArguments();
    RunId runId = ProgramRunners.getRunId(options);
    Deque<Closeable> closeables = new LinkedList<>();
    try {
        // Extract and verify parameters
        ApplicationSpecification appSpec = program.getApplicationSpecification();
        Preconditions.checkNotNull(appSpec, "Missing application specification.");
        ProgramType processorType = program.getType();
        Preconditions.checkNotNull(processorType, "Missing processor type.");
        Preconditions.checkArgument(processorType == ProgramType.SPARK, "Only Spark process type is supported.");
        SparkSpecification spec = appSpec.getSpark().get(program.getName());
        Preconditions.checkNotNull(spec, "Missing SparkSpecification for %s", program.getName());
        String host = options.getArguments().getOption(ProgramOptionConstants.HOST);
        Preconditions.checkArgument(host != null, "No hostname is provided");
        // Get the WorkflowProgramInfo if it is started by Workflow
        WorkflowProgramInfo workflowInfo = WorkflowProgramInfo.create(arguments);
        DatasetFramework programDatasetFramework = workflowInfo == null ? datasetFramework : NameMappedDatasetFramework.createFromWorkflowProgramInfo(datasetFramework, workflowInfo, appSpec);
        // Setup dataset framework context, if required
        if (programDatasetFramework instanceof ProgramContextAware) {
            ProgramId programId = program.getId();
            ((ProgramContextAware) programDatasetFramework).setContext(new BasicProgramContext(programId.run(runId)));
        }
        PluginInstantiator pluginInstantiator = createPluginInstantiator(options, program.getClassLoader());
        if (pluginInstantiator != null) {
            closeables.addFirst(pluginInstantiator);
        }
        SparkRuntimeContext runtimeContext = new SparkRuntimeContext(new Configuration(hConf), program, options, cConf, host, txClient, programDatasetFramework, metricsCollectionService, workflowInfo, pluginInstantiator, secureStore, secureStoreManager, accessEnforcer, authenticationContext, messagingService, serviceAnnouncer, pluginFinder, locationFactory, metadataReader, metadataPublisher, namespaceQueryAdmin, fieldLineageWriter, remoteClientFactory, () -> {
        });
        closeables.addFirst(runtimeContext);
        Spark spark;
        try {
            spark = new InstantiatorFactory(false).get(TypeToken.of(program.<Spark>getMainClass())).create();
        } catch (Exception e) {
            LOG.error("Failed to instantiate Spark class for {}", spec.getClassName(), e);
            throw Throwables.propagate(e);
        }
        boolean isLocal = SparkRuntimeContextConfig.isLocal(options);
        SparkSubmitter submitter;
        // If MasterEnvironment is not available, use non-master env spark submitters
        MasterEnvironment masterEnv = MasterEnvironments.getMasterEnvironment();
        if (masterEnv != null && cConf.getBoolean(Constants.Environment.PROGRAM_SUBMISSION_MASTER_ENV_ENABLED, true)) {
            submitter = new MasterEnvironmentSparkSubmitter(cConf, locationFactory, host, runtimeContext, masterEnv);
        } else {
            submitter = isLocal ? new LocalSparkSubmitter() : new DistributedSparkSubmitter(hConf, locationFactory, host, runtimeContext, options.getArguments().getOption(Constants.AppFabric.APP_SCHEDULER_QUEUE));
        }
        Service sparkRuntimeService = new SparkRuntimeService(cConf, spark, getPluginArchive(options), runtimeContext, submitter, locationFactory, isLocal, fieldLineageWriter, masterEnv);
        sparkRuntimeService.addListener(createRuntimeServiceListener(closeables), Threads.SAME_THREAD_EXECUTOR);
        ProgramController controller = new SparkProgramController(sparkRuntimeService, runtimeContext);
        LOG.debug("Starting Spark Job. Context: {}", runtimeContext);
        if (isLocal || UserGroupInformation.isSecurityEnabled()) {
            sparkRuntimeService.start();
        } else {
            ProgramRunners.startAsUser(cConf.get(Constants.CFG_HDFS_USER), sparkRuntimeService);
        }
        return controller;
    } catch (Throwable t) {
        closeAllQuietly(closeables);
        throw Throwables.propagate(t);
    }
}
Also used : ApplicationSpecification(io.cdap.cdap.api.app.ApplicationSpecification) MasterEnvironmentSparkSubmitter(io.cdap.cdap.app.runtime.spark.submit.MasterEnvironmentSparkSubmitter) MasterEnvironmentSparkSubmitter(io.cdap.cdap.app.runtime.spark.submit.MasterEnvironmentSparkSubmitter) LocalSparkSubmitter(io.cdap.cdap.app.runtime.spark.submit.LocalSparkSubmitter) SparkSubmitter(io.cdap.cdap.app.runtime.spark.submit.SparkSubmitter) DistributedSparkSubmitter(io.cdap.cdap.app.runtime.spark.submit.DistributedSparkSubmitter) Configuration(org.apache.hadoop.conf.Configuration) CConfiguration(io.cdap.cdap.common.conf.CConfiguration) Closeable(java.io.Closeable) DistributedSparkSubmitter(io.cdap.cdap.app.runtime.spark.submit.DistributedSparkSubmitter) DatasetFramework(io.cdap.cdap.data2.dataset2.DatasetFramework) NameMappedDatasetFramework(io.cdap.cdap.internal.app.runtime.workflow.NameMappedDatasetFramework) InstantiatorFactory(io.cdap.cdap.common.lang.InstantiatorFactory) SparkSpecification(io.cdap.cdap.api.spark.SparkSpecification) ProgramType(io.cdap.cdap.proto.ProgramType) RunId(org.apache.twill.api.RunId) ProgramController(io.cdap.cdap.app.runtime.ProgramController) Arguments(io.cdap.cdap.app.runtime.Arguments) MessagingService(io.cdap.cdap.messaging.MessagingService) Service(com.google.common.util.concurrent.Service) MetricsCollectionService(io.cdap.cdap.api.metrics.MetricsCollectionService) ProgramId(io.cdap.cdap.proto.id.ProgramId) BasicProgramContext(io.cdap.cdap.internal.app.runtime.BasicProgramContext) LinkedList(java.util.LinkedList) IOException(java.io.IOException) WorkflowProgramInfo(io.cdap.cdap.internal.app.runtime.workflow.WorkflowProgramInfo) MasterEnvironment(io.cdap.cdap.master.spi.environment.MasterEnvironment) PluginInstantiator(io.cdap.cdap.internal.app.runtime.plugin.PluginInstantiator) Spark(io.cdap.cdap.api.spark.Spark) LocalSparkSubmitter(io.cdap.cdap.app.runtime.spark.submit.LocalSparkSubmitter) ProgramContextAware(io.cdap.cdap.data.ProgramContextAware)

Example 7 with SparkSpecification

use of io.cdap.cdap.api.spark.SparkSpecification in project cdap by cdapio.

the class DistributedSparkProgramRunner method validateOptions.

@Override
protected void validateOptions(Program program, ProgramOptions options) {
    super.validateOptions(program, options);
    // Extract and verify parameters
    ApplicationSpecification appSpec = program.getApplicationSpecification();
    Preconditions.checkNotNull(appSpec, "Missing application specification for %s", program.getId());
    ProgramType processorType = program.getType();
    Preconditions.checkNotNull(processorType, "Missing processor type for %s", program.getId());
    Preconditions.checkArgument(processorType == ProgramType.SPARK, "Only SPARK process type is supported. Program type is %s for %s", processorType, program.getId());
    SparkSpecification spec = appSpec.getSpark().get(program.getName());
    Preconditions.checkNotNull(spec, "Missing SparkSpecification for %s", program.getId());
}
Also used : ApplicationSpecification(io.cdap.cdap.api.app.ApplicationSpecification) SparkSpecification(io.cdap.cdap.api.spark.SparkSpecification) ProgramType(io.cdap.cdap.proto.ProgramType)

Example 8 with SparkSpecification

use of io.cdap.cdap.api.spark.SparkSpecification in project cdap by cdapio.

the class SparkSpecificationCodec method deserialize.

@Override
public SparkSpecification deserialize(JsonElement json, Type typeOfT, JsonDeserializationContext context) throws JsonParseException {
    JsonObject jsonObj = json.getAsJsonObject();
    String className = jsonObj.get("className").getAsString();
    String name = jsonObj.get("name").getAsString();
    String description = jsonObj.get("description").getAsString();
    Map<String, Plugin> plugins = deserializeMap(jsonObj.get("plugins"), context, Plugin.class);
    String mainClassName = jsonObj.has("mainClassName") ? jsonObj.get("mainClassName").getAsString() : null;
    Set<String> datasets = deserializeSet(jsonObj.get("datasets"), context, String.class);
    Map<String, String> properties = deserializeMap(jsonObj.get("properties"), context, String.class);
    Resources clientResources = deserializeResources(jsonObj, "client", context);
    Resources driverResources = deserializeResources(jsonObj, "driver", context);
    Resources executorResources = deserializeResources(jsonObj, "executor", context);
    List<SparkHttpServiceHandlerSpecification> handlers = deserializeList(jsonObj.get("handlers"), context, SparkHttpServiceHandlerSpecification.class);
    return new SparkSpecification(className, name, description, mainClassName, datasets, properties, clientResources, driverResources, executorResources, handlers, plugins);
}
Also used : SparkSpecification(io.cdap.cdap.api.spark.SparkSpecification) JsonObject(com.google.gson.JsonObject) Resources(io.cdap.cdap.api.Resources) SparkHttpServiceHandlerSpecification(io.cdap.cdap.api.spark.SparkHttpServiceHandlerSpecification) Plugin(io.cdap.cdap.api.plugin.Plugin)

Example 9 with SparkSpecification

use of io.cdap.cdap.api.spark.SparkSpecification in project cdap by caskdata.

the class SparkSpecificationCodec method deserialize.

@Override
public SparkSpecification deserialize(JsonElement json, Type typeOfT, JsonDeserializationContext context) throws JsonParseException {
    JsonObject jsonObj = json.getAsJsonObject();
    String className = jsonObj.get("className").getAsString();
    String name = jsonObj.get("name").getAsString();
    String description = jsonObj.get("description").getAsString();
    Map<String, Plugin> plugins = deserializeMap(jsonObj.get("plugins"), context, Plugin.class);
    String mainClassName = jsonObj.has("mainClassName") ? jsonObj.get("mainClassName").getAsString() : null;
    Set<String> datasets = deserializeSet(jsonObj.get("datasets"), context, String.class);
    Map<String, String> properties = deserializeMap(jsonObj.get("properties"), context, String.class);
    Resources clientResources = deserializeResources(jsonObj, "client", context);
    Resources driverResources = deserializeResources(jsonObj, "driver", context);
    Resources executorResources = deserializeResources(jsonObj, "executor", context);
    List<SparkHttpServiceHandlerSpecification> handlers = deserializeList(jsonObj.get("handlers"), context, SparkHttpServiceHandlerSpecification.class);
    return new SparkSpecification(className, name, description, mainClassName, datasets, properties, clientResources, driverResources, executorResources, handlers, plugins);
}
Also used : SparkSpecification(io.cdap.cdap.api.spark.SparkSpecification) JsonObject(com.google.gson.JsonObject) Resources(io.cdap.cdap.api.Resources) SparkHttpServiceHandlerSpecification(io.cdap.cdap.api.spark.SparkHttpServiceHandlerSpecification) Plugin(io.cdap.cdap.api.plugin.Plugin)

Example 10 with SparkSpecification

use of io.cdap.cdap.api.spark.SparkSpecification in project cdap by caskdata.

the class DefaultSparkConfigurer method createSpecification.

public SparkSpecification createSpecification() {
    Set<String> datasets = new HashSet<>();
    // Grab all @Property and @Dataset fields
    Reflections.visit(spark, spark.getClass(), new PropertyFieldExtractor(properties), new DataSetFieldExtractor(datasets));
    return new SparkSpecification(spark.getClass().getName(), name, description, mainClassName, datasets, properties, clientResources, driverResources, executorResources, getHandlers(), getPlugins());
}
Also used : SparkSpecification(io.cdap.cdap.api.spark.SparkSpecification) PropertyFieldExtractor(io.cdap.cdap.internal.specification.PropertyFieldExtractor) DataSetFieldExtractor(io.cdap.cdap.internal.specification.DataSetFieldExtractor) HashSet(java.util.HashSet)

Aggregations

SparkSpecification (io.cdap.cdap.api.spark.SparkSpecification)24 ApplicationSpecification (io.cdap.cdap.api.app.ApplicationSpecification)8 ImmutableList (com.google.common.collect.ImmutableList)4 ThreadFactoryBuilder (com.google.common.util.concurrent.ThreadFactoryBuilder)4 MapReduceSpecification (io.cdap.cdap.api.mapreduce.MapReduceSpecification)4 ServiceSpecification (io.cdap.cdap.api.service.ServiceSpecification)4 Spark (io.cdap.cdap.api.spark.Spark)4 LocalizeResource (io.cdap.cdap.internal.app.runtime.distributed.LocalizeResource)4 PluginInstantiator (io.cdap.cdap.internal.app.runtime.plugin.PluginInstantiator)4 ProgramType (io.cdap.cdap.proto.ProgramType)4 ProgramId (io.cdap.cdap.proto.id.ProgramId)4 JsonObject (com.google.gson.JsonObject)3 Plugin (io.cdap.cdap.api.plugin.Plugin)3 Function (com.google.common.base.Function)2 Joiner (com.google.common.base.Joiner)2 Predicate (com.google.common.base.Predicate)2 Predicates (com.google.common.base.Predicates)2 Strings (com.google.common.base.Strings)2 Iterables (com.google.common.collect.Iterables)2 AbstractFuture (com.google.common.util.concurrent.AbstractFuture)2