List of usage examples for org.apache.hadoop.yarn.api.records ApplicationReport getDiagnostics
@Public @Stable public abstract String getDiagnostics();
From source file:com.toy.Client.java
License:Apache License
private boolean monitorApplication(ApplicationId appId) throws Exception { while (true) { // Check app status every second. try {//from w w w .jav a 2 s . co m Thread.sleep(1000); } catch (InterruptedException e) { LOG.debug("Thread sleep in monitoring loop interrupted"); } // Get application report for the appId we are interested in ApplicationReport report = yarn.getApplicationReport(appId); LOG.info("Got application report from ASM for" + ", appId=" + appId.getId() + ", appDiagnostics=" + report.getDiagnostics() + ", yarnAppState=" + report.getYarnApplicationState().toString() + ", distributedFinalState=" + report.getFinalApplicationStatus().toString() + ", appTrackingUrl=" + report.getTrackingUrl() + ", appUser=" + report.getUser()); YarnApplicationState state = report.getYarnApplicationState(); switch (state) { case RUNNING: LOG.info( "TOY is running\n" + "To stop : ./toy.sh -stop -zookeeper {}\n" + "Get containers status : ./toy.sh -status -zookeeper {}", toyConfig.zookeeper, toyConfig.zookeeper); return true; case FAILED: case KILLED: LOG.error("Ooops something went wrong"); return false; case FINISHED: LOG.info("YARN app has already finished"); return false; default: break; } } }
From source file:com.zqh.hadoop.moya.core.yarn.Client.java
License:Apache License
/** * Monitor the submitted application for completion. Kill application if * time expires./*from w w w.j ava 2 s. c om*/ * * @param appId * Application Id of application to be monitored * @return true if application completed successfully * @throws org.apache.hadoop.yarn.exceptions.YarnException * @throws java.io.IOException */ private boolean monitorApplication(ApplicationId appId) throws YarnException, IOException { while (true) { // Check app status every 1 second. try { Thread.sleep(1000); } catch (InterruptedException e) { LOG.debug("Thread sleep in monitoring loop interrupted"); } // Get application report for the appId we are interested in ApplicationReport report = yarnClient.getApplicationReport(appId); LOG.info("Got application report from ASM for" + ", appId=" + appId.getId() + ", clientToAMToken=" + report.getClientToAMToken() + ", appDiagnostics=" + report.getDiagnostics() + ", appMasterHost=" + report.getHost() + ", appQueue=" + report.getQueue() + ", appMasterRpcPort=" + report.getRpcPort() + ", appStartTime=" + report.getStartTime() + ", yarnAppState=" + report.getYarnApplicationState().toString() + ", distributedFinalState=" + report.getFinalApplicationStatus().toString() + ", appTrackingUrl=" + report.getTrackingUrl() + ", appUser=" + report.getUser()); YarnApplicationState state = report.getYarnApplicationState(); FinalApplicationStatus dsStatus = report.getFinalApplicationStatus(); if (YarnApplicationState.FINISHED == state) { if (FinalApplicationStatus.SUCCEEDED == dsStatus) { LOG.info("Application has completed successfully. Breaking monitoring loop"); return true; } else { LOG.info("Application did finished unsuccessfully." + " YarnState=" + state.toString() + ", DSFinalStatus=" + dsStatus.toString() + ". Breaking monitoring loop"); return false; } } else if (YarnApplicationState.KILLED == state || YarnApplicationState.FAILED == state) { LOG.info("Application did not finish." + " YarnState=" + state.toString() + ", DSFinalStatus=" + dsStatus.toString() + ". Breaking monitoring loop"); return false; } } }
From source file:de.huberlin.wbi.hiway.common.Client.java
License:Apache License
/** * Monitor the submitted application for completion. Kill application if time expires. * //from w w w . j av a 2 s . c o m * @param appId * Application Id of application to be monitored * @return true if application completed successfully */ private boolean monitorApplication(ApplicationId appId) throws YarnException, IOException { while (true) { // Check app status every 1 second. try { Thread.sleep(1000); } catch (InterruptedException e) { System.out.println("Thread sleep in monitoring run interrupted"); } // Get application report for the appId we are interested in ApplicationReport report = yarnClient.getApplicationReport(appId); YarnApplicationState state = report.getYarnApplicationState(); FinalApplicationStatus dsStatus = report.getFinalApplicationStatus(); if (YarnApplicationState.FINISHED == state) { if (FinalApplicationStatus.SUCCEEDED == dsStatus) { System.out.println("Application has completed successfully. Breaking monitoring run"); System.out.println(report.getDiagnostics()); return true; } System.out.println("Application finished unsuccessfully." + " YarnState=" + state.toString() + ", DSFinalStatus=" + dsStatus.toString() + ". Breaking monitoring run"); return false; } else if (YarnApplicationState.KILLED == state || YarnApplicationState.FAILED == state) { System.out.println("Application did not finish." + " YarnState=" + state.toString() + ", DSFinalStatus=" + dsStatus.toString() + ". Breaking monitoring run"); return false; } if (System.currentTimeMillis() > (clientStartTime + clientTimeout)) { System.out.println("Reached client specified timeout for application. Killing application"); forceKillApplication(appId); return false; } } }
From source file:eu.stratosphere.yarn.Client.java
License:Apache License
public void run(String[] args) throws Exception { if (UserGroupInformation.isSecurityEnabled()) { throw new RuntimeException("Stratosphere YARN client does not have security support right now." + "File a bug, we will fix it asap"); }/*from w w w. jav a 2 s. c om*/ //Utils.logFilesInCurrentDirectory(LOG); // // Command Line Options // Options options = new Options(); options.addOption(VERBOSE); options.addOption(STRATOSPHERE_CONF_DIR); options.addOption(STRATOSPHERE_JAR); options.addOption(JM_MEMORY); options.addOption(TM_MEMORY); options.addOption(TM_CORES); options.addOption(CONTAINER); options.addOption(GEN_CONF); options.addOption(QUEUE); options.addOption(QUERY); options.addOption(SHIP_PATH); CommandLineParser parser = new PosixParser(); CommandLine cmd = null; try { cmd = parser.parse(options, args); } catch (MissingOptionException moe) { System.out.println(moe.getMessage()); printUsage(); System.exit(1); } if (System.getProperty("log4j.configuration") == null) { Logger root = Logger.getRootLogger(); root.removeAllAppenders(); PatternLayout layout = new PatternLayout("%d{HH:mm:ss,SSS} %-5p %-60c %x - %m%n"); ConsoleAppender appender = new ConsoleAppender(layout, "System.err"); root.addAppender(appender); if (cmd.hasOption(VERBOSE.getOpt())) { root.setLevel(Level.DEBUG); LOG.debug("CLASSPATH: " + System.getProperty("java.class.path")); } else { root.setLevel(Level.INFO); } } // Jar Path Path localJarPath; if (cmd.hasOption(STRATOSPHERE_JAR.getOpt())) { String userPath = cmd.getOptionValue(STRATOSPHERE_JAR.getOpt()); if (!userPath.startsWith("file://")) { userPath = "file://" + userPath; } localJarPath = new Path(userPath); } else { localJarPath = new Path( "file://" + Client.class.getProtectionDomain().getCodeSource().getLocation().getPath()); } if (cmd.hasOption(GEN_CONF.getOpt())) { LOG.info("Placing default configuration in current directory"); File outFile = generateDefaultConf(localJarPath); LOG.info("File written to " + outFile.getAbsolutePath()); System.exit(0); } // Conf Path Path confPath = null; String confDirPath = ""; if (cmd.hasOption(STRATOSPHERE_CONF_DIR.getOpt())) { confDirPath = cmd.getOptionValue(STRATOSPHERE_CONF_DIR.getOpt()) + "/"; File confFile = new File(confDirPath + CONFIG_FILE_NAME); if (!confFile.exists()) { LOG.fatal("Unable to locate configuration file in " + confFile); System.exit(1); } confPath = new Path(confFile.getAbsolutePath()); } else { System.out.println("No configuration file has been specified"); // no configuration path given. // -> see if there is one in the current directory File currDir = new File("."); File[] candidates = currDir.listFiles(new FilenameFilter() { @Override public boolean accept(final File dir, final String name) { return name != null && name.endsWith(".yaml"); } }); if (candidates == null || candidates.length == 0) { System.out.println( "No configuration file has been found in current directory.\n" + "Copying default."); File outFile = generateDefaultConf(localJarPath); confPath = new Path(outFile.toURI()); } else { if (candidates.length > 1) { System.out.println("Multiple .yaml configuration files were found in the current directory\n" + "Please specify one explicitly"); System.exit(1); } else if (candidates.length == 1) { confPath = new Path(candidates[0].toURI()); } } } List<File> shipFiles = new ArrayList<File>(); // path to directory to ship if (cmd.hasOption(SHIP_PATH.getOpt())) { String shipPath = cmd.getOptionValue(SHIP_PATH.getOpt()); File shipDir = new File(shipPath); if (shipDir.isDirectory()) { shipFiles = new ArrayList<File>(Arrays.asList(shipDir.listFiles(new FilenameFilter() { @Override public boolean accept(File dir, String name) { return !(name.equals(".") || name.equals("..")); } }))); } else { LOG.warn("Ship directory is not a directory!"); } } boolean hasLog4j = false; //check if there is a log4j file if (confDirPath.length() > 0) { File l4j = new File(confDirPath + "/log4j.properties"); if (l4j.exists()) { shipFiles.add(l4j); hasLog4j = true; } } // queue String queue = "default"; if (cmd.hasOption(QUEUE.getOpt())) { queue = cmd.getOptionValue(QUEUE.getOpt()); } // JobManager Memory int jmMemory = 512; if (cmd.hasOption(JM_MEMORY.getOpt())) { jmMemory = Integer.valueOf(cmd.getOptionValue(JM_MEMORY.getOpt())); } // Task Managers memory int tmMemory = 1024; if (cmd.hasOption(TM_MEMORY.getOpt())) { tmMemory = Integer.valueOf(cmd.getOptionValue(TM_MEMORY.getOpt())); } // Task Managers vcores int tmCores = 1; if (cmd.hasOption(TM_CORES.getOpt())) { tmCores = Integer.valueOf(cmd.getOptionValue(TM_CORES.getOpt())); } Utils.getStratosphereConfiguration(confPath.toUri().getPath()); int jmPort = GlobalConfiguration.getInteger(ConfigConstants.JOB_MANAGER_IPC_PORT_KEY, 0); if (jmPort == 0) { LOG.warn("Unable to find job manager port in configuration!"); jmPort = ConfigConstants.DEFAULT_JOB_MANAGER_IPC_PORT; } conf = Utils.initializeYarnConfiguration(); // intialize HDFS LOG.info("Copy App Master jar from local filesystem and add to local environment"); // Copy the application master jar to the filesystem // Create a local resource to point to the destination jar path final FileSystem fs = FileSystem.get(conf); if (fs.getScheme().startsWith("file")) { LOG.warn("The file system scheme is '" + fs.getScheme() + "'. This indicates that the " + "specified Hadoop configuration path is wrong and the sytem is using the default Hadoop configuration values." + "The Stratosphere YARN client needs to store its files in a distributed file system"); } // Create yarnClient final YarnClient yarnClient = YarnClient.createYarnClient(); yarnClient.init(conf); yarnClient.start(); // Query cluster for metrics if (cmd.hasOption(QUERY.getOpt())) { showClusterMetrics(yarnClient); } if (!cmd.hasOption(CONTAINER.getOpt())) { LOG.fatal("Missing required argument " + CONTAINER.getOpt()); printUsage(); yarnClient.stop(); System.exit(1); } // TM Count final int taskManagerCount = Integer.valueOf(cmd.getOptionValue(CONTAINER.getOpt())); System.out.println("Using values:"); System.out.println("\tContainer Count = " + taskManagerCount); System.out.println("\tJar Path = " + localJarPath.toUri().getPath()); System.out.println("\tConfiguration file = " + confPath.toUri().getPath()); System.out.println("\tJobManager memory = " + jmMemory); System.out.println("\tTaskManager memory = " + tmMemory); System.out.println("\tTaskManager cores = " + tmCores); // Create application via yarnClient YarnClientApplication app = yarnClient.createApplication(); GetNewApplicationResponse appResponse = app.getNewApplicationResponse(); Resource maxRes = appResponse.getMaximumResourceCapability(); if (tmMemory > maxRes.getMemory() || tmCores > maxRes.getVirtualCores()) { LOG.fatal("The cluster does not have the requested resources for the TaskManagers available!\n" + "Maximum Memory: " + maxRes.getMemory() + ", Maximum Cores: " + tmCores); yarnClient.stop(); System.exit(1); } if (jmMemory > maxRes.getMemory()) { LOG.fatal("The cluster does not have the requested resources for the JobManager available!\n" + "Maximum Memory: " + maxRes.getMemory()); yarnClient.stop(); System.exit(1); } int totalMemoryRequired = jmMemory + tmMemory * taskManagerCount; ClusterResourceDescription freeClusterMem = getCurrentFreeClusterResources(yarnClient); if (freeClusterMem.totalFreeMemory < totalMemoryRequired) { LOG.fatal("This YARN session requires " + totalMemoryRequired + "MB of memory in the cluster. " + "There are currently only " + freeClusterMem.totalFreeMemory + "MB available."); yarnClient.stop(); System.exit(1); } if (tmMemory > freeClusterMem.containerLimit) { LOG.fatal("The requested amount of memory for the TaskManagers (" + tmMemory + "MB) is more than " + "the largest possible YARN container: " + freeClusterMem.containerLimit); yarnClient.stop(); System.exit(1); } if (jmMemory > freeClusterMem.containerLimit) { LOG.fatal("The requested amount of memory for the JobManager (" + jmMemory + "MB) is more than " + "the largest possible YARN container: " + freeClusterMem.containerLimit); yarnClient.stop(); System.exit(1); } // respect custom JVM options in the YAML file final String javaOpts = GlobalConfiguration.getString(ConfigConstants.STRATOSPHERE_JVM_OPTIONS, ""); // Set up the container launch context for the application master ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class); String amCommand = "$JAVA_HOME/bin/java" + " -Xmx" + Utils.calculateHeapSize(jmMemory) + "M " + javaOpts; if (hasLog4j) { amCommand += " -Dlog.file=\"" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/jobmanager-log4j.log\" -Dlog4j.configuration=file:log4j.properties"; } amCommand += " eu.stratosphere.yarn.ApplicationMaster" + " " + " 1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/jobmanager-stdout.log" + " 2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/jobmanager-stderr.log"; amContainer.setCommands(Collections.singletonList(amCommand)); System.err.println("amCommand=" + amCommand); // Set-up ApplicationSubmissionContext for the application ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); final ApplicationId appId = appContext.getApplicationId(); // Setup jar for ApplicationMaster LocalResource appMasterJar = Records.newRecord(LocalResource.class); LocalResource stratosphereConf = Records.newRecord(LocalResource.class); Path remotePathJar = Utils.setupLocalResource(conf, fs, appId.toString(), localJarPath, appMasterJar, fs.getHomeDirectory()); Path remotePathConf = Utils.setupLocalResource(conf, fs, appId.toString(), confPath, stratosphereConf, fs.getHomeDirectory()); Map<String, LocalResource> localResources = new HashMap<String, LocalResource>(2); localResources.put("stratosphere.jar", appMasterJar); localResources.put("stratosphere-conf.yaml", stratosphereConf); // setup security tokens (code from apache storm) final Path[] paths = new Path[3 + shipFiles.size()]; StringBuffer envShipFileList = new StringBuffer(); // upload ship files for (int i = 0; i < shipFiles.size(); i++) { File shipFile = shipFiles.get(i); LocalResource shipResources = Records.newRecord(LocalResource.class); Path shipLocalPath = new Path("file://" + shipFile.getAbsolutePath()); paths[3 + i] = Utils.setupLocalResource(conf, fs, appId.toString(), shipLocalPath, shipResources, fs.getHomeDirectory()); localResources.put(shipFile.getName(), shipResources); envShipFileList.append(paths[3 + i]); if (i + 1 < shipFiles.size()) { envShipFileList.append(','); } } paths[0] = remotePathJar; paths[1] = remotePathConf; paths[2] = new Path(fs.getHomeDirectory(), ".stratosphere/" + appId.toString() + "/"); FsPermission permission = new FsPermission(FsAction.ALL, FsAction.ALL, FsAction.ALL); fs.setPermission(paths[2], permission); // set permission for path. Utils.setTokensFor(amContainer, paths, this.conf); amContainer.setLocalResources(localResources); fs.close(); // Setup CLASSPATH for ApplicationMaster Map<String, String> appMasterEnv = new HashMap<String, String>(); Utils.setupEnv(conf, appMasterEnv); // set configuration values appMasterEnv.put(Client.ENV_TM_COUNT, String.valueOf(taskManagerCount)); appMasterEnv.put(Client.ENV_TM_CORES, String.valueOf(tmCores)); appMasterEnv.put(Client.ENV_TM_MEMORY, String.valueOf(tmMemory)); appMasterEnv.put(Client.STRATOSPHERE_JAR_PATH, remotePathJar.toString()); appMasterEnv.put(Client.ENV_APP_ID, appId.toString()); appMasterEnv.put(Client.ENV_CLIENT_HOME_DIR, fs.getHomeDirectory().toString()); appMasterEnv.put(Client.ENV_CLIENT_SHIP_FILES, envShipFileList.toString()); appMasterEnv.put(Client.ENV_CLIENT_USERNAME, UserGroupInformation.getCurrentUser().getShortUserName()); amContainer.setEnvironment(appMasterEnv); // Set up resource type requirements for ApplicationMaster Resource capability = Records.newRecord(Resource.class); capability.setMemory(jmMemory); capability.setVirtualCores(1); appContext.setApplicationName("Stratosphere"); // application name appContext.setAMContainerSpec(amContainer); appContext.setResource(capability); appContext.setQueue(queue); // file that we write into the conf/ dir containing the jobManager address. final File addrFile = new File(confDirPath + CliFrontend.JOBMANAGER_ADDRESS_FILE); Runtime.getRuntime().addShutdownHook(new Thread() { @Override public void run() { try { LOG.info("Killing the Stratosphere-YARN application."); yarnClient.killApplication(appId); LOG.info("Deleting files in " + paths[2]); FileSystem shutFS = FileSystem.get(conf); shutFS.delete(paths[2], true); // delete conf and jar file. shutFS.close(); } catch (Exception e) { LOG.warn("Exception while killing the YARN application", e); } try { addrFile.delete(); } catch (Exception e) { LOG.warn("Exception while deleting the jobmanager address file", e); } LOG.info("YARN Client is shutting down"); yarnClient.stop(); } }); LOG.info("Submitting application master " + appId); yarnClient.submitApplication(appContext); ApplicationReport appReport = yarnClient.getApplicationReport(appId); YarnApplicationState appState = appReport.getYarnApplicationState(); boolean told = false; char[] el = { '/', '|', '\\', '-' }; int i = 0; while (appState != YarnApplicationState.FINISHED && appState != YarnApplicationState.KILLED && appState != YarnApplicationState.FAILED) { if (!told && appState == YarnApplicationState.RUNNING) { System.err .println("Stratosphere JobManager is now running on " + appReport.getHost() + ":" + jmPort); System.err.println("JobManager Web Interface: " + appReport.getTrackingUrl()); // write jobmanager connect information PrintWriter out = new PrintWriter(addrFile); out.println(appReport.getHost() + ":" + jmPort); out.close(); addrFile.setReadable(true, false); // readable for all. told = true; } if (!told) { System.err.print(el[i++] + "\r"); if (i == el.length) { i = 0; } Thread.sleep(500); // wait for the application to switch to RUNNING } else { Thread.sleep(5000); } appReport = yarnClient.getApplicationReport(appId); appState = appReport.getYarnApplicationState(); } LOG.info("Application " + appId + " finished with" + " state " + appState + " at " + appReport.getFinishTime()); if (appState == YarnApplicationState.FAILED || appState == YarnApplicationState.KILLED) { LOG.warn("Application failed. Diagnostics " + appReport.getDiagnostics()); } }
From source file:gobblin.yarn.GobblinYarnAppLauncher.java
License:Apache License
@Subscribe public void handleApplicationReportArrivalEvent(ApplicationReportArrivalEvent applicationReportArrivalEvent) { ApplicationReport applicationReport = applicationReportArrivalEvent.getApplicationReport(); YarnApplicationState appState = applicationReport.getYarnApplicationState(); LOGGER.info("Gobblin Yarn application state: " + appState.toString()); // Reset the count on failures to get the ApplicationReport when there's one success this.getApplicationReportFailureCount.set(0); if (appState == YarnApplicationState.FINISHED || appState == YarnApplicationState.FAILED || appState == YarnApplicationState.KILLED) { applicationCompleted = true;// w w w .ja v a 2s. co m LOGGER.info("Gobblin Yarn application finished with final status: " + applicationReport.getFinalApplicationStatus().toString()); if (applicationReport.getFinalApplicationStatus() == FinalApplicationStatus.FAILED) { LOGGER.error("Gobblin Yarn application failed for the following reason: " + applicationReport.getDiagnostics()); } try { GobblinYarnAppLauncher.this.stop(); } catch (IOException ioe) { LOGGER.error("Failed to close the " + GobblinYarnAppLauncher.class.getSimpleName(), ioe); } catch (TimeoutException te) { LOGGER.error("Timeout in stopping the service manager", te); } finally { if (this.emailNotificationOnShutdown) { sendEmailOnShutdown(Optional.of(applicationReport)); } } } }
From source file:io.hops.hopsworks.common.jobs.flink.AbstractYarnClusterDescriptor.java
License:Apache License
/** * This method will block until the ApplicationMaster/JobManager have been * deployed on YARN./* w w w. ja v a 2 s.c o m*/ */ protected YarnClusterClient deployInternal() throws Exception { isReadyForDeployment(); LOG.info("Using values:"); LOG.info("\tTaskManager count = {}", taskManagerCount); LOG.info("\tJobManager memory = {}", jobManagerMemoryMb); LOG.info("\tTaskManager memory = {}", taskManagerMemoryMb); final YarnClient yarnClient = getYarnClient(); // ------------------ Check if the specified queue exists -------------------- try { List<QueueInfo> queues = yarnClient.getAllQueues(); // check only if there are queues configured in yarn and for this session. if (queues.size() > 0 && this.yarnQueue != null) { boolean queueFound = false; for (QueueInfo queue : queues) { if (queue.getQueueName().equals(this.yarnQueue)) { queueFound = true; break; } } if (!queueFound) { String queueNames = ""; for (QueueInfo queue : queues) { queueNames += queue.getQueueName() + ", "; } LOG.warn("The specified queue '" + this.yarnQueue + "' does not exist. " + "Available queues: " + queueNames); } } else { LOG.debug("The YARN cluster does not have any queues configured"); } } catch (Throwable e) { LOG.warn("Error while getting queue information from YARN: " + e.getMessage()); if (LOG.isDebugEnabled()) { LOG.debug("Error details", e); } } // Create application via yarnClient final YarnClientApplication yarnApplication = yarnClient.createApplication(); GetNewApplicationResponse appResponse = yarnApplication.getNewApplicationResponse(); Map<String, String> jobSystemProperties = new HashMap<>(2); // Certificates are materialized locally so DFSClient can be set to null // LocalResources are not used by Flink, so set it null HopsUtils.copyUserKafkaCerts(services.getUserCerts(), project, username, services.getSettings().getHopsworksTmpCertDir(), services.getSettings().getHdfsTmpCertDir(), JobType.FLINK, null, null, jobSystemProperties, services.getSettings().getFlinkKafkaCertDir(), appResponse.getApplicationId().toString()); StringBuilder tmpBuilder = new StringBuilder(); for (Map.Entry<String, String> prop : jobSystemProperties.entrySet()) { String option = YarnRunner.escapeForShell("-D" + prop.getKey() + "=" + prop.getValue()); javaOptions.add(option); addHopsworksParam(option); tmpBuilder.append(prop.getKey()).append("=").append(prop.getValue()).append("@@"); } dynamicPropertiesEncoded += tmpBuilder.toString(); // ------------------ Add dynamic properties to local flinkConfiguraton ------ Map<String, String> dynProperties = getDynamicProperties(dynamicPropertiesEncoded); for (Map.Entry<String, String> dynProperty : dynProperties.entrySet()) { flinkConfiguration.setString(dynProperty.getKey(), dynProperty.getValue()); } // ------------------ Set default file system scheme ------------------------- try { org.apache.flink.core.fs.FileSystem.setDefaultScheme(flinkConfiguration); } catch (IOException e) { throw new IOException("Error while setting the default " + "filesystem scheme from configuration.", e); } // initialize file system // Copy the application master jar to the filesystem // Create a local resource to point to the destination jar path final FileSystem fs = FileSystem.get(conf); // hard coded check for the GoogleHDFS client because its not overriding the // getScheme() method. if (!fs.getClass().getSimpleName().equals("GoogleHadoopFileSystem") && fs.getScheme().startsWith("file")) { LOG.warn("The file system scheme is '" + fs.getScheme() + "'. This indicates that the " + "specified Hadoop configuration path is wrong and the system is " + "using the default Hadoop configuration values. The Flink YARN " + "client needs to store its files in a distributed file system"); } // ------ Check if the YARN ClusterClient has the requested resources --- // the yarnMinAllocationMB specifies the smallest possible container // allocation size. all allocations below this value are automatically // set to this value. final int yarnMinAllocationMB = conf.getInt("yarn.scheduler.minimum-allocation-mb", 0); if (jobManagerMemoryMb < yarnMinAllocationMB || taskManagerMemoryMb < yarnMinAllocationMB) { LOG.warn("The JobManager or TaskManager memory is below the smallest possible " + "YARN Container size. The value of 'yarn.scheduler.minimum-allocation-mb'" + " is " + yarnMinAllocationMB + "'. Please increase the memory size." + "YARN will allocate the smaller containers but the scheduler will" + " account for the minimum-allocation-mb, maybe not all instances " + "you requested will start."); } // set the memory to minAllocationMB to do the next checks correctly if (jobManagerMemoryMb < yarnMinAllocationMB) { jobManagerMemoryMb = yarnMinAllocationMB; } if (taskManagerMemoryMb < yarnMinAllocationMB) { taskManagerMemoryMb = yarnMinAllocationMB; } Resource maxRes = appResponse.getMaximumResourceCapability(); final String NOTE = "Please check the 'yarn.scheduler.maximum-allocation-mb' and the " + "'yarn.nodemanager.resource.memory-mb' configuration values\n"; if (jobManagerMemoryMb > maxRes.getMemory()) { failSessionDuringDeployment(yarnClient, yarnApplication); throw new YarnDeploymentException("The cluster does not have the requested resources for the JobManager" + " available!\n" + "Maximum Memory: " + maxRes.getMemory() + "MB Requested: " + jobManagerMemoryMb + "MB. " + NOTE); } if (taskManagerMemoryMb > maxRes.getMemory()) { failSessionDuringDeployment(yarnClient, yarnApplication); throw new YarnDeploymentException( "The cluster does not have the requested resources for the TaskManagers available!\n" + "Maximum Memory: " + maxRes.getMemory() + " Requested: " + taskManagerMemoryMb + "MB. " + NOTE); } final String NOTE_RSC = "\nThe Flink YARN client will try to allocate the YARN session, " + "but maybe not all TaskManagers are connecting from the beginning " + "because the resources are currently not available in the cluster. " + "The allocation might take more time than usual because the Flink " + "YARN client needs to wait until the resources become available."; int totalMemoryRequired = jobManagerMemoryMb + taskManagerMemoryMb * taskManagerCount; ClusterResourceDescription freeClusterMem = getCurrentFreeClusterResources(yarnClient); if (freeClusterMem.totalFreeMemory < totalMemoryRequired) { LOG.warn("This YARN session requires " + totalMemoryRequired + "MB of memory in the cluster. " + "There are currently only " + freeClusterMem.totalFreeMemory + "MB available." + NOTE_RSC); } if (taskManagerMemoryMb > freeClusterMem.containerLimit) { LOG.warn("The requested amount of memory for the TaskManagers (" + taskManagerMemoryMb + "MB) is more than " + "the largest possible YARN container: " + freeClusterMem.containerLimit + NOTE_RSC); } if (jobManagerMemoryMb > freeClusterMem.containerLimit) { LOG.warn( "The requested amount of memory for the JobManager (" + jobManagerMemoryMb + "MB) is more than " + "the largest possible YARN container: " + freeClusterMem.containerLimit + NOTE_RSC); } // ----------------- check if the requested containers fit into the cluster. int[] nmFree = Arrays.copyOf(freeClusterMem.nodeManagersFree, freeClusterMem.nodeManagersFree.length); // first, allocate the jobManager somewhere. if (!allocateResource(nmFree, jobManagerMemoryMb)) { LOG.warn("Unable to find a NodeManager that can fit the JobManager/Application master. " + "The JobManager requires " + jobManagerMemoryMb + "MB. NodeManagers available: " + Arrays.toString(freeClusterMem.nodeManagersFree) + NOTE_RSC); } // allocate TaskManagers for (int i = 0; i < taskManagerCount; i++) { if (!allocateResource(nmFree, taskManagerMemoryMb)) { LOG.warn("There is not enough memory available in the YARN cluster. " + "The TaskManager(s) require " + taskManagerMemoryMb + "MB each. " + "NodeManagers available: " + Arrays.toString(freeClusterMem.nodeManagersFree) + "\n" + "After allocating the JobManager (" + jobManagerMemoryMb + "MB) and (" + i + "/" + taskManagerCount + ") TaskManagers, " + "the following NodeManagers are available: " + Arrays.toString(nmFree) + NOTE_RSC); } } Set<File> effectiveShipFiles = new HashSet<>(shipFiles.size()); for (File file : shipFiles) { effectiveShipFiles.add(file.getAbsoluteFile()); } //check if there is a logback or log4j file File logbackFile = new File(configurationDirectory + File.separator + CONFIG_FILE_LOGBACK_NAME); final boolean hasLogback = logbackFile.exists(); if (hasLogback) { effectiveShipFiles.add(logbackFile); } File log4jFile = new File(configurationDirectory + File.separator + CONFIG_FILE_LOG4J_NAME); final boolean hasLog4j = log4jFile.exists(); if (hasLog4j) { effectiveShipFiles.add(log4jFile); if (hasLogback) { // this means there is already a logback configuration file --> fail LOG.warn("The configuration directory ('" + configurationDirectory + "') contains both LOG4J and " + "Logback configuration files. Please delete or rename one of them."); } } addLibFolderToShipFiles(effectiveShipFiles); final ContainerLaunchContext amContainer = setupApplicationMasterContainer(hasLogback, hasLog4j); // Set-up ApplicationSubmissionContext for the application ApplicationSubmissionContext appContext = yarnApplication.getApplicationSubmissionContext(); final ApplicationId appId = appContext.getApplicationId(); // ------------------ Add Zookeeper namespace to local flinkConfiguraton ------ String zkNamespace = getZookeeperNamespace(); // no user specified cli argument for namespace? if (zkNamespace == null || zkNamespace.isEmpty()) { // namespace defined in config? else use applicationId as default. zkNamespace = flinkConfiguration.getString(ConfigConstants.ZOOKEEPER_NAMESPACE_KEY, String.valueOf(appId)); setZookeeperNamespace(zkNamespace); } flinkConfiguration.setString(ConfigConstants.ZOOKEEPER_NAMESPACE_KEY, zkNamespace); if (RecoveryMode.isHighAvailabilityModeActivated(flinkConfiguration)) { // activate re-execution of failed applications appContext.setMaxAppAttempts(flinkConfiguration.getInteger(ConfigConstants.YARN_APPLICATION_ATTEMPTS, YarnConfiguration.DEFAULT_RM_AM_MAX_ATTEMPTS)); activateHighAvailabilitySupport(appContext); } else { // set number of application retries to 1 in the default case appContext .setMaxAppAttempts(flinkConfiguration.getInteger(ConfigConstants.YARN_APPLICATION_ATTEMPTS, 1)); } // local resource map for Yarn final Map<String, LocalResource> localResources = new HashMap<>(2 + effectiveShipFiles.size()); // list of remote paths (after upload) final List<Path> paths = new ArrayList<>(2 + effectiveShipFiles.size()); // classpath assembler final StringBuilder classPathBuilder = new StringBuilder(); // ship list that enables reuse of resources for task manager containers StringBuilder envShipFileList = new StringBuilder(); // upload and register ship files for (File shipFile : effectiveShipFiles) { LocalResource shipResources = Records.newRecord(LocalResource.class); Path shipLocalPath = new Path("file://" + shipFile.getAbsolutePath()); Path remotePath = Utils.setupLocalResource(fs, appId.toString(), shipLocalPath, shipResources, fs.getHomeDirectory()); paths.add(remotePath); localResources.put(shipFile.getName(), shipResources); classPathBuilder.append(shipFile.getName()); if (shipFile.isDirectory()) { // add directories to the classpath classPathBuilder.append(File.separator).append("*"); } classPathBuilder.append(File.pathSeparator); envShipFileList.append(remotePath).append(","); } //////////////////////////////////////////////////////////////////////////// /* * Add Hops LocalResources paths here * */ //Add it to localResources for (Entry<String, LocalResource> entry : hopsworksResources.entrySet()) { localResources.put(entry.getKey(), entry.getValue()); //Append name to classPathBuilder classPathBuilder.append(entry.getKey()); classPathBuilder.append(File.pathSeparator); } //////////////////////////////////////////////////////////////////////////// // Setup jar for ApplicationMaster LocalResource appMasterJar = Records.newRecord(LocalResource.class); LocalResource flinkConf = Records.newRecord(LocalResource.class); Path remotePathJar = Utils.setupLocalResource(fs, appId.toString(), flinkJarPath, appMasterJar, fs.getHomeDirectory()); Path remotePathConf = Utils.setupLocalResource(fs, appId.toString(), flinkConfigurationPath, flinkConf, fs.getHomeDirectory()); localResources.put("flink.jar", appMasterJar); localResources.put("flink-conf.yaml", flinkConf); paths.add(remotePathJar); classPathBuilder.append("flink.jar").append(File.pathSeparator); paths.add(remotePathConf); classPathBuilder.append("flink-conf.yaml").append(File.pathSeparator); sessionFilesDir = new Path(fs.getHomeDirectory(), ".flink/" + appId.toString() + "/"); FsPermission permission = new FsPermission(FsAction.ALL, FsAction.NONE, FsAction.NONE); fs.setPermission(sessionFilesDir, permission); // set permission for path. // setup security tokens Utils.setTokensFor(amContainer, paths, conf); amContainer.setLocalResources(localResources); fs.close(); // Setup CLASSPATH and environment variables for ApplicationMaster final Map<String, String> appMasterEnv = new HashMap<>(); // set user specified app master environment variables appMasterEnv.putAll(Utils.getEnvironmentVariables(ConfigConstants.YARN_APPLICATION_MASTER_ENV_PREFIX, flinkConfiguration)); // set Flink app class path appMasterEnv.put(YarnConfigKeys.ENV_FLINK_CLASSPATH, classPathBuilder.toString()); // set Flink on YARN internal configuration values appMasterEnv.put(YarnConfigKeys.ENV_TM_COUNT, String.valueOf(taskManagerCount)); appMasterEnv.put(YarnConfigKeys.ENV_TM_MEMORY, String.valueOf(taskManagerMemoryMb)); appMasterEnv.put(YarnConfigKeys.FLINK_JAR_PATH, remotePathJar.toString()); appMasterEnv.put(YarnConfigKeys.ENV_APP_ID, appId.toString()); appMasterEnv.put(YarnConfigKeys.ENV_CLIENT_HOME_DIR, fs.getHomeDirectory().toString()); appMasterEnv.put(YarnConfigKeys.ENV_CLIENT_SHIP_FILES, envShipFileList.toString()); appMasterEnv.put(YarnConfigKeys.ENV_CLIENT_USERNAME, UserGroupInformation.getCurrentUser().getShortUserName()); appMasterEnv.put(YarnConfigKeys.ENV_SLOTS, String.valueOf(slots)); appMasterEnv.put(YarnConfigKeys.ENV_DETACHED, String.valueOf(detached)); appMasterEnv.put(YarnConfigKeys.ENV_ZOOKEEPER_NAMESPACE, getZookeeperNamespace()); if (dynamicPropertiesEncoded != null) { appMasterEnv.put(YarnConfigKeys.ENV_DYNAMIC_PROPERTIES, dynamicPropertiesEncoded); } // set classpath from YARN configuration Utils.setupYarnClassPath(conf, appMasterEnv); amContainer.setEnvironment(appMasterEnv); // Set up resource type requirements for ApplicationMaster Resource capability = Records.newRecord(Resource.class); capability.setMemory(jobManagerMemoryMb); capability.setVirtualCores(1); String name; if (customName == null) { name = "Flink session with " + taskManagerCount + " TaskManagers"; if (detached) { name += " (detached)"; } } else { name = customName; } appContext.setApplicationName(name); // application name appContext.setApplicationType("Apache Flink"); appContext.setAMContainerSpec(amContainer); appContext.setResource(capability); if (yarnQueue != null) { appContext.setQueue(yarnQueue); } // add a hook to clean up in case deployment fails Thread deploymentFailureHook = new DeploymentFailureHook(yarnClient, yarnApplication); Runtime.getRuntime().addShutdownHook(deploymentFailureHook); LOG.info("Submitting application master " + appId); yarnClient.submitApplication(appContext); LOG.info("Waiting for the cluster to be allocated"); final long startTime = System.currentTimeMillis(); ApplicationReport report; YarnApplicationState lastAppState = YarnApplicationState.NEW; loop: while (true) { try { report = yarnClient.getApplicationReport(appId); } catch (IOException e) { throw new YarnDeploymentException("Failed to deploy the cluster: " + e.getMessage()); } YarnApplicationState appState = report.getYarnApplicationState(); switch (appState) { case FAILED: case FINISHED: case KILLED: throw new YarnDeploymentException("The YARN application unexpectedly switched to state " + appState + " during deployment. \n" + "Diagnostics from YARN: " + report.getDiagnostics() + "\n" + "If log aggregation is enabled on your cluster, use this " + "command to further investigate the issue:\n" + "yarn logs -applicationId " + appId); //break .. case RUNNING: LOG.info("YARN application has been deployed successfully."); break loop; default: if (appState != lastAppState) { LOG.info("Deploying cluster, current state " + appState); } if (System.currentTimeMillis() - startTime > 60000) { LOG.info("Deployment took more than 60 seconds. Please check if the " + "requested resources are available in the YARN cluster"); } } lastAppState = appState; Thread.sleep(250); } // print the application id for user to cancel themselves. if (isDetachedMode()) { LOG.info("The Flink YARN client has been started in detached mode. In order to stop " + "Flink on YARN, use the following command or a YARN web interface to stop " + "it:\nyarn application -kill " + appId + "\nPlease also note that the " + "temporary files of the YARN session in the home directoy will not be removed."); } // since deployment was successful, remove the hook try { Runtime.getRuntime().removeShutdownHook(deploymentFailureHook); } catch (IllegalStateException e) { // we're already in the shut down hook. } String host = report.getHost(); int port = report.getRpcPort(); // Correctly initialize the Flink config flinkConfiguration.setString(ConfigConstants.JOB_MANAGER_IPC_ADDRESS_KEY, host); flinkConfiguration.setInteger(ConfigConstants.JOB_MANAGER_IPC_PORT_KEY, port); // the Flink cluster is deployed in YARN. Represent cluster return createYarnClusterClient(this, yarnClient, report, flinkConfiguration, sessionFilesDir, true); }
From source file:io.hops.hopsworks.common.jobs.flink.YarnClusterClient.java
License:Apache License
public ApplicationStatus getApplicationStatus() { if (!isConnected) { throw new IllegalStateException("The cluster has been connected to the ApplicationMaster."); }// ww w. ja v a 2s . c o m ApplicationReport lastReport = null; if (pollingRunner == null) { LOG.warn("YarnClusterClient.getApplicationStatus() has been called on an uninitialized cluster." + "The system might be in an erroneous state"); } else { lastReport = pollingRunner.getLastReport(); } if (lastReport == null) { LOG.warn("YarnClusterClient.getApplicationStatus() has been called on a " + "cluster that didn't receive a status so far." + "The system might be in an erroneous state"); return ApplicationStatus.UNKNOWN; } else { YarnApplicationState appState = lastReport.getYarnApplicationState(); ApplicationStatus status = (appState == YarnApplicationState.FAILED || appState == YarnApplicationState.KILLED) ? ApplicationStatus.FAILED : ApplicationStatus.SUCCEEDED; if (status != ApplicationStatus.SUCCEEDED) { LOG.warn("YARN reported application state {}", appState); LOG.warn("Diagnostics: {}", lastReport.getDiagnostics()); } return status; } }
From source file:io.hops.hopsworks.common.jobs.flink.YarnClusterClient.java
License:Apache License
/** * Shuts down the Yarn application/* www.j a va2s . co m*/ */ public void shutdownCluster() { if (hasBeenShutDown.getAndSet(true)) { return; } if (!isConnected) { throw new IllegalStateException("The cluster has been not been connected to the ApplicationMaster."); } try { Runtime.getRuntime().removeShutdownHook(clientShutdownHook); } catch (IllegalStateException e) { // we are already in the shutdown hook } LOG.info("Sending shutdown request to the Application Master"); try { Future<Object> response = Patterns.ask(applicationClient.get(), new YarnMessages.LocalStopYarnSession( getApplicationStatus(), "Flink YARN Client requested shutdown"), new Timeout(akkaDuration)); Await.ready(response, akkaDuration); } catch (Exception e) { LOG.warn("Error while stopping YARN cluster.", e); } try { File propertiesFile = FlinkYarnSessionCli.getYarnPropertiesLocation(flinkConfig); if (propertiesFile.isFile()) { if (propertiesFile.delete()) { LOG.info("Deleted Yarn properties file at {}", propertiesFile.getAbsoluteFile().toString()); } else { LOG.warn("Couldn't delete Yarn properties file at {}", propertiesFile.getAbsoluteFile().toString()); } } } catch (Exception e) { LOG.warn("Exception while deleting the JobManager address file", e); } if (sessionFilesDir != null) { LOG.info("Deleting files in " + sessionFilesDir); try { FileSystem shutFS = FileSystem.get(hadoopConfig); shutFS.delete(sessionFilesDir, true); // delete conf and jar file. shutFS.close(); } catch (IOException e) { LOG.error("Could not delete the Flink jar and configuration files in HDFS..", e); } } else { LOG.warn("Session file directory not set. Not deleting session files"); } try { pollingRunner.stopRunner(); pollingRunner.join(1000); } catch (InterruptedException e) { LOG.warn("Shutdown of the polling runner was interrupted", e); Thread.currentThread().interrupt(); } try { ApplicationReport appReport = yarnClient.getApplicationReport(appId); LOG.info("Application " + appId + " finished with state " + appReport.getYarnApplicationState() + " and final state " + appReport.getFinalApplicationStatus() + " at " + appReport.getFinishTime()); if (appReport.getYarnApplicationState() == YarnApplicationState.FAILED || appReport.getYarnApplicationState() == YarnApplicationState.KILLED) { LOG.warn("Application failed. Diagnostics " + appReport.getDiagnostics()); LOG.warn("If log aggregation is activated in the Hadoop cluster, we recommend to retrieve " + "the full application log using this command:\n" + "\tyarn logs -appReport " + appReport.getApplicationId() + "\n" + "(It sometimes takes a few seconds until the logs are aggregated)"); } } catch (Exception e) { LOG.warn("Couldn't get final report", e); } LOG.info("YARN Client is shutting down"); yarnClient.stop(); // actorRunner is using the yarnClient. yarnClient = null; // set null to clearly see if somebody wants to access it afterwards. }
From source file:io.hops.tensorflow.Client.java
License:Apache License
/** * Monitor the submitted application for completion. * Kill application if time expires./*w w w.ja v a2 s.c om*/ * * @param appId * Application Id of application to be monitored * @return true if application completed successfully * @throws YarnException * @throws IOException */ public boolean monitorApplication(ApplicationId appId) throws YarnException, IOException { YarnApplicationState oldState = null; while (true) { // Check app status every 1 second. try { Thread.sleep(1000); } catch (InterruptedException e) { LOG.debug("Thread sleep in monitoring loop interrupted"); } ApplicationReport report = yarnClient.getApplicationReport(appId); YarnApplicationState state = report.getYarnApplicationState(); FinalApplicationStatus dsStatus = report.getFinalApplicationStatus(); if (oldState != state) { LOG.info("Got application report from ASM for" + "\n\t appId=" + appId.getId() + "\n\t clientToAMToken=" + report.getClientToAMToken() + "\n\t appDiagnostics=" + report.getDiagnostics() + "\n\t appMasterHost=" + report.getHost() + "\n\t appQueue=" + report.getQueue() + "\n\t appMasterRpcPort=" + report.getRpcPort() + "\n\t appStartTime=" + report.getStartTime() + "\n\t yarnAppState=" + report.getYarnApplicationState().toString() + "\n\t distributedFinalState=" + report.getFinalApplicationStatus().toString() + "\n\t appTrackingUrl=" + report.getTrackingUrl() + "\n\t appUser=" + report.getUser()); oldState = state; } else { LOG.info("Got application report from ASM for " + appId + " (state: " + state + ")"); } if (YarnApplicationState.FINISHED == state) { if (FinalApplicationStatus.SUCCEEDED == dsStatus) { LOG.info("Application has completed successfully. Breaking monitoring loop"); return true; } else { LOG.info("Application did finished unsuccessfully." + " YarnState=" + state.toString() + ", DSFinalStatus=" + dsStatus.toString() + ". Breaking monitoring loop"); return false; } } else if (YarnApplicationState.KILLED == state || YarnApplicationState.FAILED == state) { LOG.info("Application did not finish." + " YarnState=" + state.toString() + ", DSFinalStatus=" + dsStatus.toString() + ". Breaking monitoring loop"); return false; } if (System.currentTimeMillis() > (clientStartTime + clientTimeout)) { LOG.info("Reached client specified timeout for application. Killing application"); forceKillApplication(appId); return false; } } }
From source file:MasteringYarn.DistributedShellClient.java
public void run(String[] args) throws YarnException, IOException, InterruptedException { YarnConfiguration yarnConfiguration = new YarnConfiguration(); YarnClient yarnClient = YarnClient.createYarnClient(); yarnClient.init(yarnConfiguration);/* ww w. ja va2 s. c o m*/ yarnClient.start(); YarnClientApplication yarnClientApplication = yarnClient.createApplication(); //container launch context for application master ContainerLaunchContext applicationMasterContainer = Records.newRecord(ContainerLaunchContext.class); applicationMasterContainer.setCommands( Collections.singletonList("$JAVA_HOME/bin/java MasteringYarn.DistributedShellApplicationMaster " + args[2] + " " + args[3] + " " + "1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stdout " + "2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stderr")); LocalResource applicationMasterJar = Records.newRecord(LocalResource.class); setupJarFileForApplicationMaster(new Path(args[1]), applicationMasterJar); applicationMasterContainer .setLocalResources(Collections.singletonMap("MasteringYarn.jar", applicationMasterJar)); Map<String, String> appMasterEnv = new HashMap<>(); setupEnvironmentForApplicationMaster(appMasterEnv); applicationMasterContainer.setEnvironment(appMasterEnv); Resource resources = Records.newRecord(Resource.class); resources.setVirtualCores(1); resources.setMemory(100); ApplicationSubmissionContext submissionContext = yarnClientApplication.getApplicationSubmissionContext(); submissionContext.setAMContainerSpec(applicationMasterContainer); submissionContext.setQueue("default"); submissionContext.setApplicationName("MasteringYarn"); submissionContext.setResource(resources); ApplicationId applicationId = submissionContext.getApplicationId(); System.out.println("Submitting " + applicationId); yarnClient.submitApplication(submissionContext); System.out.println("Post submission " + applicationId); ApplicationReport applicationReport; YarnApplicationState applicationState; do { Thread.sleep(1000); applicationReport = yarnClient.getApplicationReport(applicationId); applicationState = applicationReport.getYarnApplicationState(); System.out.println("Diagnostics " + applicationReport.getDiagnostics()); } while (applicationState != YarnApplicationState.FAILED && applicationState != YarnApplicationState.FINISHED && applicationState != YarnApplicationState.KILLED); System.out.println("Application finished with " + applicationState + " state and id " + applicationId); }