List of usage examples for org.apache.hadoop.yarn.api.protocolrecords RegisterApplicationMasterResponse getContainersFromPreviousAttempts
@Public @Unstable public abstract List<Container> getContainersFromPreviousAttempts();
Get the list of running containers as viewed by ResourceManager
from previous application attempts.
From source file:ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/*from ww w.j a v a2 s .c o m*/ * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException { LOG.info("Starting ApplicationMaster"); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_START); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.CallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); // Setup local RPC Server to accept status requests directly from clients // TODO need to setup a protocol for client to be able to communicate to // the RPC server // TODO use the rpc port info to register with the RM for the client to // send requests to this app master // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capabililty of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_END); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } }
From source file:cn.edu.buaa.act.petuumOnYarn.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException//from w ww . j a v a 2 s. co m * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException { LOG.info("Starting ApplicationMaster"); // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer // class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.CallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); // Setup local RPC Server to accept status requests directly from // clients // TODO need to setup a protocol for client to be able to communicate to // the RPC server // TODO use the rpc port info to register with the RM for the client to // send requests to this app master // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capabililty of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers; // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); }
From source file:com.github.hdl.tensorflow.yarn.app.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException//from w w w . java 2 s .co m * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException, InterruptedException { LOG.info("Starting ApplicationMaster"); // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.AbstractCallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); appMasterHostname = System.getenv(Environment.NM_HOST.name()); TFApplicationRpcServer rpcServer = new TFApplicationRpcServer(appMasterHostname, new RpcForClient()); appMasterRpcPort = rpcServer.getRpcPort(); rpcServer.startRpcServiceThread(); // Register self with ResourceManager // This will start heartbeating to the RM RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager long maxMem = response.getMaximumResourceCapability().getMemorySize(); LOG.info("Max mem capability of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capability of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); for (Container container : previousAMRunningContainers) { launchedContainers.add(container.getId()); } numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); }
From source file:com.inforefiner.hdata.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/*from w w w . j av a 2s . co m*/ * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException, InterruptedException { LOG.info("Starting ApplicationMaster"); // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.CallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); startTimelineClient(conf); if (timelineClient != null) { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_START, domainId, appSubmitterUgi); } // Setup local RPC Server to accept status requests directly from clients // TODO need to setup a protocol for client to be able to communicate to // the RPC server // TODO use the rpc port info to register with the RM for the client to // send requests to this app master // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capabililty of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); }
From source file:com.scistor.dshell.ScistorApplicationMaster.java
License:Apache License
/** * Main run function for the application master * //from w w w . j a v a 2 s. c o m * @throws YarnException * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException { LOG.info("Starting ApplicationMaster"); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_START); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.CallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); // Setup local RPC Server to accept status requests directly from // clients // TODO need to setup a protocol for client to be able to communicate to // the RPC server // TODO use the rpc port info to register with the RM for the client to // send requests to this app master // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capabililty of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info("Received " + previousAMRunningContainers.size() + " previous AM's running containers on AM registration."); numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainersToRequest); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_END); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } }
From source file:com.sogou.dockeronyarn.service.DockerApplicationMaster_24.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/*from w w w . j a v a 2 s . com*/ * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException { LOG.info("Starting ApplicationMaster"); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_START); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.CallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); // Setup local RPC Server to accept status requests directly from clients // TODO need to setup a protocol for client to be able to communicate to // the RPC server // TODO use the rpc port info to register with the RM for the client to // send requests to this app master // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capabililty of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_END); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } numRetryCount.set(0); }
From source file:io.hops.tensorflow.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/* w w w . j av a 2 s . c o m*/ * @throws IOException */ public void run() throws YarnException, IOException, InterruptedException { LOG.info("Starting ApplicationMaster. " + "Workers: " + numWorkers + ", Parameter servers: " + numPses); clusterSpecServer = new ClusterSpecGeneratorServer(appAttemptID.getApplicationId().toString(), numTotalContainers, numWorkers); LOG.info("Starting ClusterSpecGeneratorServer"); int port = 2222; while (true) { try { clusterSpecServer.start(port); break; } catch (IOException e) { port++; } } environment.put("YARNTF_AM_ADDRESS", InetAddress.getLocalHost().getHostName() + ":" + port); environment.put("YARNTF_APPLICATION_ID", appAttemptID.getApplicationId().toString()); // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); rmWrapper = new RMWrapper(this); rmWrapper.getClient().init(conf); rmWrapper.getClient().start(); nmWrapper = new NMWrapper(this); nmWrapper.getClient().init(conf); nmWrapper.getClient().start(); timelineHandler = new TimelineHandler(appAttemptID.toString(), domainId, appSubmitterUgi); timelineHandler.startClient(conf); if (timelineHandler.isClientNotNull()) { timelineHandler.publishApplicationAttemptEvent(YarntfEvent.YARNTF_APP_ATTEMPT_START); } // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); appMasterTrackingUrl = InetAddress.getLocalHost().getHostName() + ":" + TensorBoardServer.spawn(this); RegisterApplicationMasterResponse response = rmWrapper.getClient() .registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capabililty of resources in this cluster " + maxVCores); int maxGPUS = response.getMaximumResourceCapability().getGPUs(); LOG.info("Max gpus capabililty of resources in this cluster " + maxGPUS); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } if (containerGPUs > maxGPUS) { LOG.info("Container gpus specified above max threshold of cluster." + " Using max value." + ", specified=" + containerGPUs + ", max=" + maxGPUS); containerGPUs = maxGPUS; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); // Stop eventual containers from previous attempts for (Container prevContainer : previousAMRunningContainers) { LOG.info("Releasing YARN container " + prevContainer.getId()); rmWrapper.getClient().releaseAssignedContainer(prevContainer.getId()); } // Send request for containers to RM for (int i = 0; i < numWorkers; i++) { ContainerRequest workerRequest = setupContainerAskForRM(true); rmWrapper.getClient().addContainerRequest(workerRequest); } numRequestedContainers.addAndGet(numWorkers); for (int i = 0; i < numPses; i++) { ContainerRequest psRequest = setupContainerAskForRM(false); rmWrapper.getClient().addContainerRequest(psRequest); } numRequestedContainers.addAndGet(numPses); }
From source file:me.haosdent.noya.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws org.apache.hadoop.yarn.exceptions.YarnException * @throws java.io.IOException//from w w w . j av a2 s . c o m */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException { LOG.info("Starting ApplicationMaster"); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_START); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.CallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); // Setup local RPC Server to accept status requests directly from clients // TODO need to setup a protocol for client to be able to communicate to // the RPC server // TODO use the rpc port info to register with the RM for the client to // send requests to this app master // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capabililty of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info("Received " + previousAMRunningContainers.size() + " previous AM's running containers on AM registration."); numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainersToRequest); try { publishApplicationAttemptEvent(timelineClient, appAttemptID.toString(), DSEvent.DS_APP_ATTEMPT_END); } catch (Exception e) { LOG.error("App Attempt start event coud not be pulished for " + appAttemptID.toString(), e); } }
From source file:org.apache.hama.bsp.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * //from w w w . ja v a2 s. c o m * @throws org.apache.hadoop.yarn.exceptions.YarnException * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException, InterruptedException { LOG.info("Starting ApplicationMaster"); // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.CallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(localConf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(localConf); nmClientAsync.start(); // Setup local RPC Server to accept status requests directly from clients // TODO need to setup a protocol for client to be able to communicate to // the RPC server // TODO use the rpc port info to register with the RM for the client to // send requests to this app master // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capability of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capability of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); for (Container container : previousAMRunningContainers) { launchedContainers.add(container.getId()); } numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { AMRMClient.ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); }
From source file:org.hdl.caffe.yarn.app.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/*w ww .ja va 2 s . co m*/ * @throws IOException */ @SuppressWarnings({ "unchecked" }) public void run() throws YarnException, IOException, InterruptedException { LOG.info("Starting ApplicationMaster"); // Note: Credentials, Token, UserGroupInformation, DataOutputBuffer class // are marked as LimitedPrivate Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); // Now remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); LOG.info("Executing with tokens:"); while (iter.hasNext()) { Token<?> token = iter.next(); LOG.info(token); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Create appSubmitterUgi and add original tokens to it String appSubmitterUserName = System.getenv(ApplicationConstants.Environment.USER.name()); appSubmitterUgi = UserGroupInformation.createRemoteUser(appSubmitterUserName); appSubmitterUgi.addCredentials(credentials); AMRMClientAsync.AbstractCallbackHandler allocListener = new RMCallbackHandler(); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); appMasterHostname = System.getenv(Environment.NM_HOST.name()); CaffeApplicationRpcServer rpcServer = new CaffeApplicationRpcServer(appMasterHostname, new RpcForClient()); appMasterRpcPort = rpcServer.getRpcPort(); rpcServer.startRpcServiceThread(); // Register self with ResourceManager // This will start heartbeating to the RM RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // Dump out information about cluster capability as seen by the // resource manager long maxMem = response.getMaximumResourceCapability().getMemorySize(); LOG.info("Max mem capability of resources in this cluster " + maxMem); int maxVCores = response.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max vcores capability of resources in this cluster " + maxVCores); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { LOG.info("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerVirtualCores > maxVCores) { LOG.info("Container virtual cores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerVirtualCores + ", max=" + maxVCores); containerVirtualCores = maxVCores; } List<Container> previousAMRunningContainers = response.getContainersFromPreviousAttempts(); LOG.info(appAttemptID + " received " + previousAMRunningContainers.size() + " previous attempts' running containers on AM registration."); for (Container container : previousAMRunningContainers) { launchedContainers.add(container.getId()); } numAllocatedContainers.addAndGet(previousAMRunningContainers.size()); int numTotalContainersToRequest = numTotalContainers - previousAMRunningContainers.size(); // Setup ask for containers from RM // Send request for containers to RM // Until we get our fully allocated quota, we keep on polling RM for // containers // Keep looping until all the containers are launched and shell script // executed on them ( regardless of success/failure). for (int i = 0; i < numTotalContainersToRequest; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); }