List of usage examples for org.apache.hadoop.security Credentials getAllTokens
public Collection<Token<? extends TokenIdentifier>> getAllTokens()
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 . j ava 2 s .com * @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// w w w .j av a 2s. c om * @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 * // w w w .j av a 2s . 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.client.DockerApplicationMaster_23.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/* ww w . j a v a 2s.co m*/ * @throws IOException */ @SuppressWarnings({ "unchecked" }) public boolean run() throws YarnException, IOException { LOG.info("Starting ApplicationMaster"); 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(); while (iter.hasNext()) { Token<?> token = iter.next(); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); 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; } // 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 < numTotalContainers; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } //numRequestedContainers.set(numTotalContainers); while (!done && (numCompletedContainers.get() != numTotalContainers)) { try { Thread.sleep(200); } catch (InterruptedException ex) { } } finish(); return success; }
From source file:com.sogou.dockeronyarn.service.DockerApplicationMaster_24.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException// w ww. jav a 2 s. co 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); } numRetryCount.set(0); }
From source file:com.srini.hadoopYarn.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/* w w w. j a v a2s .c o m*/ * @throws IOException */ @SuppressWarnings({ "unchecked" }) public boolean run() throws YarnException, IOException { LOG.info("Starting ApplicationMaster"); 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(); while (iter.hasNext()) { Token<?> token = iter.next(); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); 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); // 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; } // 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 < numTotalContainers; ++i) { ContainerRequest containerAsk = setupContainerAskForRM(); amRMClient.addContainerRequest(containerAsk); } numRequestedContainers.set(numTotalContainers); while (!done && (numCompletedContainers.get() != numTotalContainers)) { try { Thread.sleep(200); } catch (InterruptedException ex) { } } finish(); return success; }
From source file:com.tito.easyyarn.appmaster.ApplicationMaster.java
License:Apache License
private void extractTokens() { // Credentials, Token, UserGroupInformation, DataOutputBuffer Credentials credentials; try {/* w w w . j ava 2 s . co m*/ 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()); } catch (IOException e) { LOG.error("extractTokens error={}", e); } }
From source file:de.huberlin.wbi.hiway.am.HiWay.java
License:Apache License
/** * Main run function for the application master * /*from w w w . j av a 2 s. c o m*/ * @return True if there were no errors * @throws YarnException * YarnException * @throws IOException * IOException */ @SuppressWarnings("unchecked") public boolean run() throws YarnException, IOException { System.out.println("Starting ApplicationMaster"); Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); try (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(); while (iter.hasNext()) { Token<?> token = iter.next(); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); allocListener = new RMCallbackHandler(this); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); containerListener = new NMCallbackHandler(this); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); Data workflowData = new Data(workflowPath); workflowData.stageIn(); // Register self with ResourceManager. This will start heartbeating to the RM. appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); switch (schedulerName) { case staticRoundRobin: case heft: scheduler = schedulerName.equals(HiWayConfiguration.HIWAY_SCHEDULER_OPTS.staticRoundRobin) ? new RoundRobin(getWorkflowName(), hdfs, conf) : new HEFT(getWorkflowName(), hdfs, conf); break; case greedyQueue: scheduler = new GreedyQueue(getWorkflowName(), conf, hdfs); break; default: C3PO c3po = new C3PO(getWorkflowName(), hdfs, conf); switch (schedulerName) { case conservative: c3po.setConservatismWeight(12d); c3po.setnClones(0); c3po.setPlacementAwarenessWeight(0.01d); c3po.setOutlookWeight(0.01d); break; case cloning: c3po.setConservatismWeight(0.01d); c3po.setnClones(1); c3po.setPlacementAwarenessWeight(0.01d); c3po.setOutlookWeight(0.01d); break; case placementAware: c3po.setConservatismWeight(0.01d); c3po.setnClones(0); c3po.setPlacementAwarenessWeight(12d); c3po.setOutlookWeight(0.01d); break; case outlooking: c3po.setConservatismWeight(0.01d); c3po.setnClones(0); c3po.setPlacementAwarenessWeight(0.01d); c3po.setOutlookWeight(12d); break; default: c3po.setConservatismWeight(3d); c3po.setnClones(2); c3po.setPlacementAwarenessWeight(1d); c3po.setOutlookWeight(2d); } scheduler = c3po; } scheduler.initialize(); writeEntryToLog(new JsonReportEntry(getRunId(), null, null, null, null, null, HiwayDBI.KEY_WF_NAME, getWorkflowName())); parseWorkflow(); scheduler.updateRuntimeEstimates(getRunId().toString()); federatedReport = new Data(appId + ".log"); // Dump out information about cluster capability as seen by the resource manager int maxMem = response.getMaximumResourceCapability().getMemory(); int maxCores = response.getMaximumResourceCapability().getVirtualCores(); System.out.println("Max mem capabililty of resources in this cluster " + maxMem); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { System.out.println("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerCores > maxCores) { System.out.println("Container vcores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerCores + ", max=" + maxCores); containerCores = maxCores; } while (!done) { try { while (scheduler.hasNextNodeRequest()) { ContainerRequest containerAsk = setupContainerAskForRM(scheduler.getNextNodeRequest()); amRMClient.addContainerRequest(containerAsk); } Thread.sleep(1000); System.out.println("Current application state: requested=" + numRequestedContainers + ", completed=" + numCompletedContainers + ", failed=" + numFailedContainers + ", killed=" + numKilledContainers + ", allocated=" + numAllocatedContainers); } catch (InterruptedException e) { e.printStackTrace(); System.exit(-1); } } finish(); } catch (Exception e) { e.printStackTrace(); System.exit(-1); } return success; }
From source file:de.huberlin.wbi.hiway.am.WorkflowDriver.java
License:Apache License
/** * Main run function for the application master. Does more initialization (sic!). * Calls the abstract {@link #parseWorkflow()}, then {@link #executeWorkflow()} and finally {@link #finish()}. * @return True if there were no errors//from w ww .java 2 s .c o m */ protected boolean run() throws IOException { /* log */ Logger.writeToStdout("Starting ApplicationMaster"); Credentials credentials = UserGroupInformation.getCurrentUser().getCredentials(); try (DataOutputBuffer dob = new DataOutputBuffer()) { credentials.writeTokenStorageToStream(dob); // remove the AM->RM token so that containers cannot access it. Iterator<Token<?>> iter = credentials.getAllTokens().iterator(); while (iter.hasNext()) { Token<?> token = iter.next(); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); // Resource Manager communications setup RMCallbackHandler allocListener = new RMCallbackHandler(this); amRMClient = AMRMClientAsync.createAMRMClientAsync(1000, allocListener); amRMClient.init(conf); amRMClient.start(); // Node Managers communications setup containerListener = new NMCallbackHandler(this); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); // get workflow file if (hdfs.exists(workflowPath)) { Path localPath = new Path(workflowPath.getName()); hdfs.copyToLocalFile(false, workflowPath, localPath); workflowPath = localPath; workflowFile = new Data(workflowPath); workflowFile.stageOut(); } else { // TODO this doesn't work; the path is triggered when running the application e.g., as hiway workflows/test.dax // but stageIn then fails, because in the HDFS, there is only test.dax and not workflows/test.dax workflowFile = new Data(workflowPath); workflowFile.stageIn(); } // Register self with ResourceManager. This will start heartbeating to the RM. /* the hostname of the container running the Hi-WAY ApplicationMaster */ String appMasterHostname = NetUtils.getHostname(); /* the port on which the ApplicationMaster listens for status updates from clients */ int appMasterRpcPort = -1; /* the tracking URL to which the ApplicationMaster publishes info for clients to monitor */ String appMasterTrackingUrl = ""; RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); // initialize scheduler switch (schedulerEnumValue) { case roundRobin: case heft: int workerMemory = conf.getInt(YarnConfiguration.NM_PMEM_MB, YarnConfiguration.DEFAULT_NM_PMEM_MB); scheduler = schedulerEnumValue.equals(HiWayConfiguration.HIWAY_SCHEDULERS.roundRobin) ? new RoundRobin(getWorkflowName()) : new HEFT(getWorkflowName(), workerMemory / containerMemory); break; case greedy: scheduler = new GreedyQueue(getWorkflowName()); break; case memoryAware: scheduler = new MemoryAware(getWorkflowName(), amRMClient); break; case perfectDaxGQ: scheduler = new PerfectDaxGreedyQueue(getWorkflowName()); break; default: C3PO c3po = new C3PO(getWorkflowName()); switch (schedulerEnumValue) { case dataAware: c3po.setConservatismWeight(0.01d); c3po.setnClones(0); c3po.setPlacementAwarenessWeight(12d); c3po.setOutlookWeight(0.01d); break; default: c3po.setConservatismWeight(3d); c3po.setnClones(2); c3po.setPlacementAwarenessWeight(1d); c3po.setOutlookWeight(2d); } scheduler = c3po; } scheduler.init(conf, hdfs, containerMemory, customMemoryMap, containerCores, requestPriority); scheduler.initializeProvenanceManager(); /* log */ logger.writeEntryToLog(new JsonReportEntry(getRunId(), null, null, null, null, null, HiwayDBI.KEY_WF_NAME, getWorkflowName())); logger.federatedReport = new Data(appId + ".log"); // parse workflow, obtain ready tasks Collection<TaskInstance> readyTasks = parseWorkflow(); // scheduler updates runtime estimates for all tasks comprising the workflow scheduler.updateRuntimeEstimates(getRunId().toString()); scheduler.addTasks(readyTasks); // Dump out information about cluster capability as seen by the resource manager maxMem = response.getMaximumResourceCapability().getMemory(); maxCores = response.getMaximumResourceCapability().getVirtualCores(); /* log */ Logger.writeToStdout("Max mem capabililty of resources in this cluster " + maxMem); // A resource ask cannot exceed the max. if (containerMemory > maxMem) { /* log */ Logger.writeToStdout("Container memory specified above max threshold of cluster." + " Using max value." + ", specified=" + containerMemory + ", max=" + maxMem); containerMemory = maxMem; } if (containerCores > maxCores) { /* log */ Logger.writeToStdout("Container vcores specified above max threshold of cluster." + " Using max value." + ", specified=" + containerCores + ", max=" + maxCores); containerCores = maxCores; } // this is the actual work loop: // ask for resources until the workflow is done. executeWorkflow(); finish(); } catch (Exception e) { e.printStackTrace(System.out); System.exit(-1); } return success; }
From source file:edu.cmu.graphchi.toolkits.collaborative_filtering.yarn.ApplicationMaster.java
License:Apache License
/** * Main run function for the application master * * @throws YarnException/*ww w.j a va 2 s . c om*/ * @throws IOException */ @SuppressWarnings({ "unchecked" }) public boolean run() throws YarnException, IOException { yarnClient.start(); LOG.info("Starting ApplicationMaster"); 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(); while (iter.hasNext()) { Token<?> token = iter.next(); if (token.getKind().equals(AMRMTokenIdentifier.KIND_NAME)) { iter.remove(); } } allTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); amRMClient = AMRMClient.createAMRMClient(); amRMClient.init(conf); amRMClient.start(); containerListener = createNMCallbackHandler(); nmClientAsync = new NMClientAsyncImpl(containerListener); nmClientAsync.init(conf); nmClientAsync.start(); // Register self with ResourceManager // This will start heartbeating to the RM appMasterHostname = NetUtils.getHostname(); RegisterApplicationMasterResponse response = amRMClient.registerApplicationMaster(appMasterHostname, appMasterRpcPort, appMasterTrackingUrl); //TODO: Figure out how to do this. List<NodeReport> reports = this.yarnClient.getNodeReports(); LOG.info("Cluster Status"); List<Resource> availableResources = new ArrayList<Resource>(); for (NodeReport nr : reports) { LOG.info(" NodeId: " + nr.getNodeId() + " Capabilities " + nr.getCapability() + " Used Resources " + nr.getUsed()); int availableMem = nr.getCapability().getMemory() - nr.getUsed().getMemory(); int availableVCores = nr.getCapability().getVirtualCores() - nr.getUsed().getVirtualCores(); Resource resource = Resource.newInstance(availableMem, availableVCores); availableResources.add(resource); } /*Based on available resources scheduler should decide the best allocation of recommenders to resources and return a list of resources that should be requested from the ResourceManager*/ DataSetDescription datasetDesc = new DataSetDescription(this.setup.dataMetadataFile); List<RecommenderPool> recommenderPools = RecommenderScheduler.splitRecommenderPool(availableResources, recommenders, datasetDesc, this.setup.nShards); for (RecommenderPool res : recommenderPools) { ContainerRequest containerAsk = setupContainerAskForRM(res.getTotalMemory(), requestPriority); LOG.info("CONTAINER ASK: " + containerAsk); amRMClient.addContainerRequest(containerAsk); } float progress = 0; List<RecommenderPool> pendingPools = new ArrayList<RecommenderPool>(); for (RecommenderPool p : recommenderPools) pendingPools.add(p); this.numTotalContainers = recommenderPools.size(); this.numCompletedContainers.set(0); this.numAllocatedContainers.set(0); while (numCompletedContainers.get() != numTotalContainers) { try { Thread.sleep(200); AllocateResponse allocResp = amRMClient.allocate(progress); List<Container> newContainers = allocResp.getAllocatedContainers(); List<ContainerStatus> completedContainers = allocResp.getCompletedContainersStatuses(); if (this.numAllocatedContainers.get() >= this.numTotalContainers && pendingPools.size() != 0) { //Ask for new containers for pending pools LOG.warn("The number of allocated containers has exceeded number of total containers, but " + "the pending pool size is still not 0. Asking for new containers from RM"); for (RecommenderPool res : pendingPools) { ContainerRequest containerAsk = setupContainerAskForRM(res.getTotalMemory(), requestPriority); LOG.info("NEW CONTAINER ASK: " + containerAsk); amRMClient.addContainerRequest(containerAsk); } } if (newContainers.size() > 0) { LOG.info("Allocated " + newContainers.size() + " new containers"); numAllocatedContainers.addAndGet(newContainers.size()); for (Container container : newContainers) { //Find matching recommender pool from pendingRecommender pools. RecommenderPool pool = null; for (RecommenderPool p : pendingPools) { if (p.getTotalMemory() == container.getResource().getMemory()) { pool = p; break; } } if (pool == null) { LOG.warn("No Takers for Container " + container + " Releasing container"); amRMClient.releaseAssignedContainer(container.getId()); } else { startContainer(container, pool); //This pool has now got a container. Remove it from pending pools pendingPools.remove(pool); } } } onContainersCompleted(completedContainers); } catch (InterruptedException ex) { } } finish(); return success; }