Examples with ApplicationSubmissionContext org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext used on opensource projects

Example 91 with ApplicationSubmissionContext

use of org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext in project tez by apache.

the class TezClientUtils method createApplicationSubmissionContext.

/**
 * Create an ApplicationSubmissionContext to launch a Tez AM
 * @param appId Application Id
 * @param dag DAG to be submitted
 * @param amName Name for the application
 * @param amConfig AM Configuration
 * @param tezJarResources Resources to be used by the AM
 * @param sessionCreds the credential object which will be populated with session specific
 * @param servicePluginsDescriptor descriptor for services which may be running in the AM
 * @return an ApplicationSubmissionContext to launch a Tez AM
 * @throws IOException
 * @throws YarnException
 */
@Private
@VisibleForTesting
public static ApplicationSubmissionContext createApplicationSubmissionContext(ApplicationId appId, DAG dag, String amName, AMConfiguration amConfig, Map<String, LocalResource> tezJarResources, Credentials sessionCreds, boolean tezLrsAsArchive, TezApiVersionInfo apiVersionInfo, ServicePluginsDescriptor servicePluginsDescriptor, JavaOptsChecker javaOptsChecker) throws IOException, YarnException {
    Preconditions.checkNotNull(sessionCreds);
    TezConfiguration conf = amConfig.getTezConfiguration();
    FileSystem fs = TezClientUtils.ensureStagingDirExists(conf, TezCommonUtils.getTezBaseStagingPath(conf));
    String strAppId = appId.toString();
    Path tezSysStagingPath = TezCommonUtils.createTezSystemStagingPath(conf, strAppId);
    Path binaryConfPath = TezCommonUtils.getTezConfStagingPath(tezSysStagingPath);
    binaryConfPath = fs.makeQualified(binaryConfPath);
    // Setup resource requirements
    Resource capability = Records.newRecord(Resource.class);
    capability.setMemory(amConfig.getTezConfiguration().getInt(TezConfiguration.TEZ_AM_RESOURCE_MEMORY_MB, TezConfiguration.TEZ_AM_RESOURCE_MEMORY_MB_DEFAULT));
    capability.setVirtualCores(amConfig.getTezConfiguration().getInt(TezConfiguration.TEZ_AM_RESOURCE_CPU_VCORES, TezConfiguration.TEZ_AM_RESOURCE_CPU_VCORES_DEFAULT));
    if (LOG.isDebugEnabled()) {
        LOG.debug("AppMaster capability = " + capability);
    }
    // Setup required Credentials for the AM launch. DAG specific credentials
    // are handled separately.
    ByteBuffer securityTokens = null;
    // Setup security tokens
    Credentials amLaunchCredentials = new Credentials();
    if (amConfig.getCredentials() != null) {
        amLaunchCredentials.addAll(amConfig.getCredentials());
    }
    // Add Staging dir creds to the list of session credentials.
    TokenCache.obtainTokensForFileSystems(sessionCreds, new Path[] { binaryConfPath }, conf);
    // Add session specific credentials to the AM credentials.
    amLaunchCredentials.mergeAll(sessionCreds);
    DataOutputBuffer dob = new DataOutputBuffer();
    amLaunchCredentials.writeTokenStorageToStream(dob);
    securityTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength());
    // Setup the command to run the AM
    List<String> vargs = new ArrayList<String>(8);
    vargs.add(Environment.JAVA_HOME.$() + "/bin/java");
    String amOpts = constructAMLaunchOpts(amConfig.getTezConfiguration(), capability);
    vargs.add(amOpts);
    String amLogLevelString = amConfig.getTezConfiguration().get(TezConfiguration.TEZ_AM_LOG_LEVEL, TezConfiguration.TEZ_AM_LOG_LEVEL_DEFAULT);
    String[] amLogParams = parseLogParams(amLogLevelString);
    String amLogLevel = amLogParams[0];
    maybeAddDefaultLoggingJavaOpts(amLogLevel, vargs);
    // FIX sun bug mentioned in TEZ-327
    vargs.add("-Dsun.nio.ch.bugLevel=''");
    vargs.add(TezConstants.TEZ_APPLICATION_MASTER_CLASS);
    if (dag == null) {
        vargs.add("--" + TezConstants.TEZ_SESSION_MODE_CLI_OPTION);
    }
    vargs.add("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + File.separator + ApplicationConstants.STDOUT);
    vargs.add("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + File.separator + ApplicationConstants.STDERR);
    Vector<String> vargsFinal = new Vector<String>(8);
    // Final command
    StringBuilder mergedCommand = new StringBuilder();
    for (CharSequence str : vargs) {
        mergedCommand.append(str).append(" ");
    }
    vargsFinal.add(mergedCommand.toString());
    if (LOG.isDebugEnabled()) {
        LOG.debug("Command to launch container for ApplicationMaster is : " + mergedCommand);
    }
    Map<String, String> environment = new TreeMap<String, String>();
    TezYARNUtils.setupDefaultEnv(environment, conf, TezConfiguration.TEZ_AM_LAUNCH_ENV, TezConfiguration.TEZ_AM_LAUNCH_ENV_DEFAULT, TezConfiguration.TEZ_AM_LAUNCH_CLUSTER_DEFAULT_ENV, TezConfiguration.TEZ_AM_LAUNCH_CLUSTER_DEFAULT_ENV_DEFAULT, tezLrsAsArchive);
    addVersionInfoToEnv(environment, apiVersionInfo);
    addLogParamsToEnv(environment, amLogParams);
    Map<String, LocalResource> amLocalResources = new TreeMap<String, LocalResource>();
    // Not fetching credentials for AMLocalResources. Expect this to be provided via AMCredentials.
    if (amConfig.getAMLocalResources() != null) {
        amLocalResources.putAll(amConfig.getAMLocalResources());
    }
    amLocalResources.putAll(tezJarResources);
    TezConfiguration tezConf = amConfig.getTezConfiguration();
    // Merge the dag access controls into tez am config.
    if (dag != null && dag.getDagAccessControls() != null) {
        // Merge updates the conf object passed. In non session mode, same client object can be used
        // to submit multiple dags, copying this prevents ACL of one DAG from being used in another.
        tezConf = new TezConfiguration(amConfig.getTezConfiguration());
        dag.getDagAccessControls().mergeIntoAmAcls(tezConf);
    }
    // don't overwrite existing conf, needed for TezClient.getClient() so existing containers have stable resource fingerprints
    if (!binaryConfPath.getFileSystem(tezConf).exists(binaryConfPath)) {
        ConfigurationProto finalConfProto = createFinalConfProtoForApp(tezConf, servicePluginsDescriptor);
        FSDataOutputStream amConfPBOutBinaryStream = null;
        try {
            amConfPBOutBinaryStream = TezCommonUtils.createFileForAM(fs, binaryConfPath);
            finalConfProto.writeTo(amConfPBOutBinaryStream);
        } finally {
            if (amConfPBOutBinaryStream != null) {
                amConfPBOutBinaryStream.close();
            }
        }
    }
    LocalResource binaryConfLRsrc = TezClientUtils.createLocalResource(fs, binaryConfPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION);
    amConfig.setBinaryConfLR(binaryConfLRsrc);
    amLocalResources.put(TezConstants.TEZ_PB_BINARY_CONF_NAME, binaryConfLRsrc);
    // Create Session Jars definition to be sent to AM as a local resource
    Path sessionJarsPath = TezCommonUtils.getTezAMJarStagingPath(tezSysStagingPath);
    FSDataOutputStream sessionJarsPBOutStream = null;
    try {
        sessionJarsPBOutStream = TezCommonUtils.createFileForAM(fs, sessionJarsPath);
        // Write out the initial list of resources which will be available in the AM
        DAGProtos.PlanLocalResourcesProto amResourceProto;
        if (amLocalResources != null && !amLocalResources.isEmpty()) {
            amResourceProto = DagTypeConverters.convertFromLocalResources(amLocalResources);
        } else {
            amResourceProto = DAGProtos.PlanLocalResourcesProto.getDefaultInstance();
        }
        amResourceProto.writeDelimitedTo(sessionJarsPBOutStream);
    } finally {
        if (sessionJarsPBOutStream != null) {
            sessionJarsPBOutStream.close();
        }
    }
    LocalResource sessionJarsPBLRsrc = TezClientUtils.createLocalResource(fs, sessionJarsPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION);
    amLocalResources.put(TezConstants.TEZ_AM_LOCAL_RESOURCES_PB_FILE_NAME, sessionJarsPBLRsrc);
    String user = UserGroupInformation.getCurrentUser().getShortUserName();
    ACLManager aclManager = new ACLManager(user, amConfig.getTezConfiguration());
    Map<ApplicationAccessType, String> acls = aclManager.toYARNACls();
    if (dag != null) {
        DAGPlan dagPB = prepareAndCreateDAGPlan(dag, amConfig, tezJarResources, tezLrsAsArchive, sessionCreds, servicePluginsDescriptor, javaOptsChecker);
        // emit protobuf DAG file style
        Path binaryPath = TezCommonUtils.getTezBinPlanStagingPath(tezSysStagingPath);
        if (LOG.isDebugEnabled()) {
            LOG.debug("Stage directory information for AppId :" + appId + " tezSysStagingPath :" + tezSysStagingPath + " binaryConfPath :" + binaryConfPath + " sessionJarsPath :" + sessionJarsPath + " binaryPlanPath :" + binaryPath);
        }
        FSDataOutputStream dagPBOutBinaryStream = null;
        try {
            // binary output
            dagPBOutBinaryStream = TezCommonUtils.createFileForAM(fs, binaryPath);
            dagPB.writeTo(dagPBOutBinaryStream);
        } finally {
            if (dagPBOutBinaryStream != null) {
                dagPBOutBinaryStream.close();
            }
        }
        amLocalResources.put(TezConstants.TEZ_PB_PLAN_BINARY_NAME, TezClientUtils.createLocalResource(fs, binaryPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION));
        if (Level.DEBUG.isGreaterOrEqual(Level.toLevel(amLogLevel))) {
            Path textPath = localizeDagPlanAsText(dagPB, fs, amConfig, strAppId, tezSysStagingPath);
            amLocalResources.put(TezConstants.TEZ_PB_PLAN_TEXT_NAME, TezClientUtils.createLocalResource(fs, textPath, LocalResourceType.FILE, LocalResourceVisibility.APPLICATION));
        }
    }
    // Send the shuffle token as part of the AM launch context, so that the NM running the AM can
    // provide this to AuxServices running on the AM node - in case tasks run within the AM,
    // and no other task runs on this node.
    Map<String, ByteBuffer> serviceData = new HashMap<String, ByteBuffer>();
    String auxiliaryService = conf.get(TezConfiguration.TEZ_AM_SHUFFLE_AUXILIARY_SERVICE_ID, TezConfiguration.TEZ_AM_SHUFFLE_AUXILIARY_SERVICE_ID_DEFAULT);
    serviceData.put(auxiliaryService, TezCommonUtils.serializeServiceData(TokenCache.getSessionToken(amLaunchCredentials)));
    // Setup ContainerLaunchContext for AM container
    ContainerLaunchContext amContainer = ContainerLaunchContext.newInstance(amLocalResources, environment, vargsFinal, serviceData, securityTokens, acls);
    // Set up the ApplicationSubmissionContext
    ApplicationSubmissionContext appContext = Records.newRecord(ApplicationSubmissionContext.class);
    Collection<String> tagsFromConf = amConfig.getTezConfiguration().getTrimmedStringCollection(TezConfiguration.TEZ_APPLICATION_TAGS);
    appContext.setApplicationType(TezConstants.TEZ_APPLICATION_TYPE);
    if (tagsFromConf != null && !tagsFromConf.isEmpty()) {
        appContext.setApplicationTags(new HashSet<String>(tagsFromConf));
    }
    appContext.setApplicationId(appId);
    appContext.setResource(capability);
    String queueName = amConfig.getQueueName();
    if (queueName != null && !queueName.isEmpty()) {
        appContext.setQueue(amConfig.getQueueName());
    }
    // set the application priority
    setApplicationPriority(appContext, amConfig);
    appContext.setApplicationName(amName);
    appContext.setCancelTokensWhenComplete(amConfig.getTezConfiguration().getBoolean(TezConfiguration.TEZ_CANCEL_DELEGATION_TOKENS_ON_COMPLETION, TezConfiguration.TEZ_CANCEL_DELEGATION_TOKENS_ON_COMPLETION_DEFAULT));
    appContext.setAMContainerSpec(amContainer);
    appContext.setMaxAppAttempts(amConfig.getTezConfiguration().getInt(TezConfiguration.TEZ_AM_MAX_APP_ATTEMPTS, TezConfiguration.TEZ_AM_MAX_APP_ATTEMPTS_DEFAULT));
    return appContext;
}

Example 92 with ApplicationSubmissionContext

use of org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext in project tez by apache.

the class TezClient method setupApplicationContext.

private ApplicationSubmissionContext setupApplicationContext() throws IOException, YarnException {
    TezClientUtils.processTezLocalCredentialsFile(sessionCredentials, amConfig.getTezConfiguration());
    Map<String, LocalResource> tezJarResources = getTezJarResources(sessionCredentials);
    // Add session token for shuffle
    TezClientUtils.createSessionToken(sessionAppId.toString(), jobTokenSecretManager, sessionCredentials);
    ApplicationSubmissionContext appContext = TezClientUtils.createApplicationSubmissionContext(sessionAppId, null, clientName, amConfig, tezJarResources, sessionCredentials, usingTezArchiveDeploy, apiVersionInfo, servicePluginsDescriptor, javaOptsChecker);
    // Set Tez Sessions to not retry on AM crashes if recovery is disabled
    if (!amConfig.getTezConfiguration().getBoolean(TezConfiguration.DAG_RECOVERY_ENABLED, TezConfiguration.DAG_RECOVERY_ENABLED_DEFAULT)) {
        appContext.setMaxAppAttempts(1);
    }
    return appContext;
}

Example 93 with ApplicationSubmissionContext

use of org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext in project tez by apache.

the class TezClient method start.

/**
 * Start the client. This establishes a connection to the YARN cluster.
 * In session mode, this start the App Master thats runs all the DAGs in the
 * session.
 * @throws TezException
 * @throws IOException
 */
public synchronized void start() throws TezException, IOException {
    amConfig.setYarnConfiguration(new YarnConfiguration(amConfig.getTezConfiguration()));
    startFrameworkClient();
    setupJavaOptsChecker();
    if (isSession) {
        LOG.info("Session mode. Starting session.");
        TezClientUtils.processTezLocalCredentialsFile(sessionCredentials, amConfig.getTezConfiguration());
        clientTimeout = amConfig.getTezConfiguration().getInt(TezConfiguration.TEZ_SESSION_CLIENT_TIMEOUT_SECS, TezConfiguration.TEZ_SESSION_CLIENT_TIMEOUT_SECS_DEFAULT);
        try {
            if (sessionAppId == null) {
                sessionAppId = createApplication();
            }
            ApplicationSubmissionContext appContext = setupApplicationContext();
            frameworkClient.submitApplication(appContext);
            ApplicationReport appReport = frameworkClient.getApplicationReport(sessionAppId);
            LOG.info("The url to track the Tez Session: " + appReport.getTrackingUrl());
            sessionStarted.set(true);
        } catch (YarnException e) {
            throw new TezException(e);
        }
        startClientHeartbeat();
        this.stagingFs = FileSystem.get(amConfig.getTezConfiguration());
    }
}

Example 94 with ApplicationSubmissionContext

use of org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext in project tez by apache.

the class TezClient method submitDAGApplication.

// To be used only by YarnRunner
@Private
DAGClient submitDAGApplication(ApplicationId appId, DAG dag) throws TezException, IOException {
    LOG.info("Submitting DAG application with id: " + appId);
    try {
        // Use the AMCredentials object in client mode, since this won't be re-used.
        // Ensures we don't fetch credentially unnecessarily if the user has already provided them.
        Credentials credentials = amConfig.getCredentials();
        if (credentials == null) {
            credentials = new Credentials();
        }
        TezClientUtils.processTezLocalCredentialsFile(credentials, amConfig.getTezConfiguration());
        // Add session token for shuffle
        TezClientUtils.createSessionToken(appId.toString(), jobTokenSecretManager, credentials);
        // Add credentials for tez-local resources.
        Map<String, LocalResource> tezJarResources = getTezJarResources(credentials);
        ApplicationSubmissionContext appContext = TezClientUtils.createApplicationSubmissionContext(appId, dag, dag.getName(), amConfig, tezJarResources, credentials, usingTezArchiveDeploy, apiVersionInfo, servicePluginsDescriptor, javaOptsChecker);
        String callerContextStr = "";
        if (dag.getCallerContext() != null) {
            callerContextStr = ", callerContext=" + dag.getCallerContext().contextAsSimpleString();
        }
        LOG.info("Submitting DAG to YARN" + ", applicationId=" + appId + ", dagName=" + dag.getName() + callerContextStr);
        frameworkClient.submitApplication(appContext);
        ApplicationReport appReport = frameworkClient.getApplicationReport(appId);
        LOG.info("The url to track the Tez AM: " + appReport.getTrackingUrl());
        lastSubmittedAppId = appId;
    } catch (YarnException e) {
        throw new TezException(e);
    }
    // wait for dag in non-session mode to start running, so that we can start to getDAGStatus
    waitNonSessionTillReady();
    return getDAGClient(appId, amConfig.getTezConfiguration(), amConfig.getYarnConfiguration(), frameworkClient);
}

Example 95 with ApplicationSubmissionContext

use of org.apache.hadoop.yarn.api.records.ApplicationSubmissionContext in project tez by apache.

the class TestTezClient method testGetClient.

@Test(timeout = 5000)
public void testGetClient() throws Exception {
    /* BEGIN first TezClient usage without calling stop() */
    TezClientForTest client = testTezClient(true, false);
    /* END first TezClient usage without calling stop() */
    /* BEGIN reuse of AM from new TezClient */
    ArgumentCaptor<ApplicationSubmissionContext> captor = ArgumentCaptor.forClass(ApplicationSubmissionContext.class);
    when(client.mockYarnClient.getApplicationReport(client.mockAppId).getYarnApplicationState()).thenReturn(YarnApplicationState.RUNNING);
    // Reuse existing appId from first TezClient
    ApplicationId existingAppId = client.mockAppId;
    TezClientForTest client2 = configureAndCreateTezClient(null, true, client.amConfig.getTezConfiguration());
    String mockLR1Name = "LR1";
    Map<String, LocalResource> lrDAG = Collections.singletonMap(mockLR1Name, LocalResource.newInstance(URL.newInstance("file", "localhost", 0, "/test1"), LocalResourceType.FILE, LocalResourceVisibility.PUBLIC, 1, 1));
    Vertex vertex = Vertex.create("Vertex", ProcessorDescriptor.create("P"), 1, Resource.newInstance(1, 1));
    DAG dag = DAG.create("DAG").addVertex(vertex).addTaskLocalFiles(lrDAG);
    // Bind TezClient to existing app and submit a dag
    DAGClient dagClient = client2.getClient(existingAppId).submitDAG(dag);
    assertTrue(dagClient.getExecutionContext().contains(existingAppId.toString()));
    assertEquals(dagClient.getSessionIdentifierString(), existingAppId.toString());
    // Validate request for new AM is not submitted to RM */
    verify(client2.mockYarnClient, times(0)).submitApplication(captor.capture());
    // Validate dag submission from second TezClient as normal */
    verify(client2.sessionAmProxy, times(1)).submitDAG((RpcController) any(), (SubmitDAGRequestProto) any());
    // Validate stop from new TezClient as normal */
    client2.stop();
    verify(client2.sessionAmProxy, times(1)).shutdownSession((RpcController) any(), (ShutdownSessionRequestProto) any());
    verify(client2.mockYarnClient, times(1)).stop();
/* END reuse of AM from new TezClient */
}