List of usage examples for org.apache.hadoop.yarn.api.records YarnApplicationState KILLED
YarnApplicationState KILLED
To view the source code for org.apache.hadoop.yarn.api.records YarnApplicationState KILLED.
Click Source Link
From source file:io.hops.metadata.util.TestHopYarnAPIUtilities.java
License:Apache License
@Test(timeout = 30000) public void testForceKillApplication() throws Exception { MockRM rm = new MockRM(conf); rm.start();//from w w w .j a v a 2s .com ClientRMService rmService = rm.getClientRMService(); GetApplicationsRequest getRequest = GetApplicationsRequest .newInstance(EnumSet.of(YarnApplicationState.KILLED)); ApplicationId appId1 = getApplicationId(100); ApplicationACLsManager mockAclsManager = mock(ApplicationACLsManager.class); when(mockAclsManager.checkAccess(UserGroupInformation.getCurrentUser(), ApplicationAccessType.VIEW_APP, null, appId1)).thenReturn(true); SubmitApplicationRequest submitRequest1 = mockSubmitAppRequest(appId1, null, null); try { rmService.submitApplication(submitRequest1); } catch (YarnException e) { Assert.fail("Exception is not expected."); } assertEquals("Incorrect number of apps in the RM", 0, rmService.getApplications(getRequest).getApplicationList().size()); Thread.sleep(1000); //TODO: check what have to be present in the db Thread.sleep(2000); rm.stop(); Thread.sleep(2000); }
From source file:io.hops.tensorflow.Client.java
License:Apache License
/** * Monitor the submitted application for completion. * Kill application if time expires./*w ww. ja va2 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);// w w w.j a v a 2s .co 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); }
From source file:org.apache.ambari.view.slider.SliderAppsViewControllerImpl.java
License:Apache License
private SliderApp createSliderAppObject(ApplicationReport yarnApp, Set<String> properties, SliderClient sliderClient) {/* w w w .ja v a2s .c o m*/ if (yarnApp == null) { return null; } SliderApp app = new SliderApp(); app.setState(yarnApp.getYarnApplicationState().name()); // Valid Slider App? // We want all Slider apps except the ones which properly finished. if (YarnApplicationState.FINISHED.equals(yarnApp.getYarnApplicationState())) { try { if (sliderClient.actionExists(yarnApp.getName(), false) == LauncherExitCodes.EXIT_SUCCESS) { app.setState(SliderApp.STATE_FROZEN); } } catch (UnknownApplicationInstanceException e) { return null; // Application not in HDFS - means it is not frozen } catch (YarnException e) { logger.warn("Unable to determine frozen state for " + yarnApp.getName(), e); return null; } catch (IOException e) { logger.warn("Unable to determine frozen state for " + yarnApp.getName(), e); return null; } } if (YarnApplicationState.KILLED.equals(yarnApp.getYarnApplicationState()) || YarnApplicationState.FAILED.equals(yarnApp.getYarnApplicationState())) { try { if (sliderClient.actionExists(yarnApp.getName(), false) != LauncherExitCodes.EXIT_SUCCESS) { // YARN application is killed or failed, and no HDFS content - Application has been destroyed. return null; } } catch (UnknownApplicationInstanceException e) { return null; // Application not in HDFS - means it is not frozen } catch (YarnException e) { logger.warn("Unable to determine status of killed app " + yarnApp.getName(), e); return null; } catch (IOException e) { logger.warn("Unable to determine status of killed app " + yarnApp.getName(), e); return null; } } app.setId(getApplicationIdString(yarnApp.getApplicationId())); app.setName(yarnApp.getName()); app.setUser(yarnApp.getUser()); app.setDiagnostics(yarnApp.getDiagnostics()); app.setYarnId(yarnApp.getApplicationId().toString()); app.setStartTime(yarnApp.getStartTime()); app.setEndTime(yarnApp.getFinishTime()); Set<String> applicationTags = yarnApp.getApplicationTags(); if (applicationTags != null && applicationTags.size() > 0) { for (String tag : applicationTags) { int index = tag.indexOf(':'); if (index > 0 && index < tag.length() - 1) { String key = tag.substring(0, index).trim(); String value = tag.substring(index + 1).trim(); if ("name".equals(key)) { app.setType(value); } else if ("version".equals(key)) { app.setAppVersion(value); } else if ("description".equals(key)) { app.setDescription(value); } } } } if (properties != null && !properties.isEmpty()) { SliderAppType matchedAppType = null; List<SliderAppType> matchingAppTypes = getSliderAppTypes(null); if (matchingAppTypes != null && matchingAppTypes.size() > 0) { for (SliderAppType appType : matchingAppTypes) { if ((appType.getTypeName() != null && appType.getTypeName().equalsIgnoreCase(app.getType())) && (appType.getTypeVersion() != null && appType.getTypeVersion().equalsIgnoreCase(app.getAppVersion()))) { matchedAppType = appType; break; } } } SliderAppMasterClient sliderAppClient = yarnApp.getTrackingUrl() == null ? null : new SliderAppMasterClient(yarnApp.getTrackingUrl()); SliderAppMasterData appMasterData = null; Map<String, String> quickLinks = new HashMap<String, String>(); Set<String> metrics = new HashSet<String>(); for (String property : properties) { if ("RUNNING".equals(app.getState())) { if (sliderAppClient != null) { if (appMasterData == null) { appMasterData = sliderAppClient.getAppMasterData(); } if (appMasterData != null && "urls".equals(property.toLowerCase())) { if (quickLinks.isEmpty()) { quickLinks = sliderAppClient.getQuickLinks(appMasterData.publisherUrl); } app.setUrls(quickLinks); } else if (appMasterData != null && "configs".equals(property.toLowerCase())) { Map<String, Map<String, String>> configs = sliderAppClient .getConfigs(appMasterData.publisherUrl); app.setConfigs(configs); } else if (appMasterData != null && "jmx".equals(property.toLowerCase())) { if (quickLinks.isEmpty()) { quickLinks = sliderAppClient.getQuickLinks(appMasterData.publisherUrl); } if (quickLinks != null && quickLinks.containsKey("JMX")) { String jmxUrl = quickLinks.get("JMX"); if (matchedAppType != null) { MetricsHolder metricsHolder = appMetrics.get(matchedAppType.uniqueName()); app.setJmx(sliderAppClient.getJmx(jmxUrl, viewContext, matchedAppType, metricsHolder)); } } Map<String, Map<String, String>> configs = sliderAppClient .getConfigs(appMasterData.publisherUrl); app.setConfigs(configs); } else if ("components".equals(property.toLowerCase())) { try { ClusterDescription description = sliderClient .getClusterDescription(yarnApp.getName()); if (description != null && description.status != null && !description.status.isEmpty()) { Map<String, SliderAppComponent> componentTypeMap = new HashMap<String, SliderAppComponent>(); for (Entry<String, Object> e : description.status.entrySet()) { @SuppressWarnings("unchecked") Map<String, Map<String, Map<String, Object>>> componentsObj = (Map<String, Map<String, Map<String, Object>>>) e .getValue(); boolean isLive = "live".equals(e.getKey()); for (Entry<String, Map<String, Map<String, Object>>> componentEntry : componentsObj .entrySet()) { SliderAppComponent appComponent = componentTypeMap .get(componentEntry.getKey()); if (appComponent == null) { appComponent = new SliderAppComponent(); appComponent.setComponentName(componentEntry.getKey()); appComponent.setActiveContainers( new HashMap<String, Map<String, String>>()); appComponent.setCompletedContainers( new HashMap<String, Map<String, String>>()); componentTypeMap.put(componentEntry.getKey(), appComponent); } for (Entry<String, Map<String, Object>> containerEntry : componentEntry .getValue().entrySet()) { Map<String, String> containerDataMap = new HashMap<String, String>(); String containerId = containerEntry.getKey(); Map<String, Object> containerValues = containerEntry.getValue(); for (String containerProperty : containerValues.keySet()) { Object containerPropertyValue = containerValues .get(containerProperty); containerDataMap.put(containerProperty, containerPropertyValue.toString()); } if (isLive) { appComponent.getActiveContainers().put(containerId, containerDataMap); } else { appComponent.getCompletedContainers().put(containerId, containerDataMap); } } // Set total instances count from statistics appComponent.setInstanceCount(appComponent.getActiveContainers().size() + appComponent.getCompletedContainers().size()); if (description.statistics != null && description.statistics .containsKey(componentEntry.getKey())) { Map<String, Integer> statisticsMap = description.statistics .get(componentEntry.getKey()); if (statisticsMap.containsKey("containers.desired")) { appComponent.setInstanceCount( statisticsMap.get("containers.desired")); } } } } app.setAlerts( sliderAlerts.generateComponentsAlerts(componentTypeMap, app.getType())); app.setComponents(componentTypeMap); } } catch (UnknownApplicationInstanceException e) { logger.warn("Unable to determine app components for " + yarnApp.getName(), e); } catch (YarnException e) { logger.warn("Unable to determine app components for " + yarnApp.getName(), e); throw new RuntimeException(e.getMessage(), e); } catch (IOException e) { logger.warn("Unable to determine app components for " + yarnApp.getName(), e); throw new RuntimeException(e.getMessage(), e); } } else if (property.startsWith(METRICS_PREFIX)) { metrics.add(property.substring(METRICS_PREFIX.length())); } else if ("supportedMetrics".equals(property)) { if (matchedAppType != null) { app.setSupportedMetrics(matchedAppType.getSupportedMetrics()); } } } } } if (metrics.size() > 0) { if (quickLinks.isEmpty()) { quickLinks = sliderAppClient.getQuickLinks(appMasterData.publisherUrl); } if (quickLinks != null && quickLinks.containsKey(METRICS_API_NAME)) { String metricsUrl = quickLinks.get(METRICS_API_NAME); MetricsHolder metricsHolder = appMetrics.get(matchedAppType.uniqueName()); app.setMetrics(sliderAppClient.getMetrics(yarnApp.getName(), metricsUrl, metrics, null, viewContext, matchedAppType, metricsHolder)); } } } return app; }
From source file:org.apache.drill.yarn.core.YarnRMClient.java
License:Apache License
/** * Wait for the application to enter one of the completion states. This is an * informal implementation useful for testing. * * @throws YarnClientException//from www. ja v a 2 s . c o m */ public void waitForCompletion() throws YarnClientException { ApplicationReport appReport; YarnApplicationState appState; for (;;) { appReport = getAppReport(); appState = appReport.getYarnApplicationState(); if (appState == YarnApplicationState.FINISHED || appState == YarnApplicationState.KILLED || appState == YarnApplicationState.FAILED) { break; } try { Thread.sleep(100); } catch (InterruptedException e) { // Should never occur. } } System.out.println("Application " + appId + " finished with" + " state " + appState + " at " + appReport.getFinishTime()); }
From source file:org.apache.flink.yarn.cli.YarnApplicationStatusMonitor.java
License:Apache License
private void updateApplicationStatus() { if (yarnClient.isInState(Service.STATE.STARTED)) { final ApplicationReport applicationReport; try {//from w w w .j a v a2 s . co m applicationReport = yarnClient.getApplicationReport(yarnApplicationId); } catch (Exception e) { LOG.info("Could not retrieve the Yarn application report for {}.", yarnApplicationId); applicationStatus = ApplicationStatus.UNKNOWN; return; } YarnApplicationState yarnApplicationState = applicationReport.getYarnApplicationState(); if (yarnApplicationState == YarnApplicationState.FAILED || yarnApplicationState == YarnApplicationState.KILLED) { applicationStatus = ApplicationStatus.FAILED; } else { applicationStatus = ApplicationStatus.SUCCEEDED; } } else { LOG.info("Yarn client is no longer in state STARTED. Stopping the Yarn application status monitor."); applicationStatusUpdateFuture.cancel(false); } }
From source file:org.apache.flink.yarn.Client.java
License:Apache License
public void run(String[] args) throws Exception { if (UserGroupInformation.isSecurityEnabled()) { throw new RuntimeException("Flink YARN client does not have security support right now." + "File a bug, we will fix it asap"); }/*from w w w .ja v a 2s .c o m*/ //Utils.logFilesInCurrentDirectory(LOG); // // Command Line Options // Options options = new Options(); options.addOption(VERBOSE); options.addOption(FLINK_CONF_DIR); options.addOption(FLINK_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(FLINK_JAR.getOpt())) { String userPath = cmd.getOptionValue(FLINK_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(FLINK_CONF_DIR.getOpt())) { confDirPath = cmd.getOptionValue(FLINK_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.getFlinkConfiguration(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 Flink 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.FLINK_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 += " org.apache.flink.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 flinkConf = 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, flinkConf, fs.getHomeDirectory()); Map<String, LocalResource> localResources = new HashMap<String, LocalResource>(2); localResources.put("flink.jar", appMasterJar); localResources.put("flink-conf.yaml", flinkConf); // 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(), ".flink/" + 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.FLINK_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("Flink"); // 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 Flink-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("Flink 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:org.apache.flink.yarn.FlinkYarnCluster.java
License:Apache License
/** * Connect the FlinkYarnCluster to the ApplicationMaster. * * Detached YARN sessions don't need to connect to the ApplicationMaster. * Detached per job YARN sessions need to connect until the required number of TaskManagers have been started. * /* ww w.ja v a 2 s .c o m*/ * @throws IOException */ public void connectToCluster() throws IOException { if (isConnected) { throw new IllegalStateException("Can not connect to the cluster again"); } // start actor system LOG.info("Start actor system."); // find name of own public interface, able to connect to the JM // try to find address for 2 seconds. log after 400 ms. InetAddress ownHostname = ConnectionUtils.findConnectingAddress(jobManagerAddress, 2000, 400); actorSystem = AkkaUtils.createActorSystem(flinkConfig, new Some<Tuple2<String, Object>>( new Tuple2<String, Object>(ownHostname.getCanonicalHostName(), 0))); // Create the leader election service flinkConfig.setString(ConfigConstants.JOB_MANAGER_IPC_ADDRESS_KEY, this.jobManagerAddress.getHostName()); flinkConfig.setInteger(ConfigConstants.JOB_MANAGER_IPC_PORT_KEY, this.jobManagerAddress.getPort()); LeaderRetrievalService leaderRetrievalService; try { leaderRetrievalService = LeaderRetrievalUtils.createLeaderRetrievalService(flinkConfig); } catch (Exception e) { throw new IOException("Could not create the leader retrieval service.", e); } // start application client LOG.info("Start application client."); applicationClient = actorSystem.actorOf( Props.create(ApplicationClient.class, flinkConfig, leaderRetrievalService), "applicationClient"); actorRunner = new Thread(new Runnable() { @Override public void run() { // blocks until ApplicationClient has been stopped actorSystem.awaitTermination(); // get final application report 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 -applicationId " + appReport.getApplicationId() + "\n" + "(It sometimes takes a few seconds until the logs are aggregated)"); } } catch (Exception e) { LOG.warn("Error while getting final application report", e); } } }); actorRunner.setDaemon(true); actorRunner.start(); pollingRunner = new PollingThread(yarnClient, appId); pollingRunner.setDaemon(true); pollingRunner.start(); Runtime.getRuntime().addShutdownHook(clientShutdownHook); isConnected = true; }
From source file:org.apache.flink.yarn.FlinkYarnCluster.java
License:Apache License
@Override public boolean hasFailed() { if (!isConnected) { throw new IllegalStateException("The cluster has been connected to the ApplicationMaster."); }// w ww .java 2 s. c o m if (pollingRunner == null) { LOG.warn("FlinkYarnCluster.hasFailed() has been called on an uninitialized cluster." + "The system might be in an erroneous state"); } ApplicationReport lastReport = pollingRunner.getLastReport(); if (lastReport == null) { LOG.warn( "FlinkYarnCluster.hasFailed() has been called on a cluster that didn't receive a status so far." + "The system might be in an erroneous state"); return false; } else { YarnApplicationState appState = lastReport.getYarnApplicationState(); boolean status = (appState == YarnApplicationState.FAILED || appState == YarnApplicationState.KILLED); if (status) { LOG.warn("YARN reported application state {}", appState); LOG.warn("Diagnostics: {}", lastReport.getDiagnostics()); } return status; } }
From source file:org.apache.flink.yarn.YarnClusterClient.java
License:Apache License
public ApplicationStatus getApplicationStatus() { if (!isConnected) { throw new IllegalStateException("The cluster has been connected to the ApplicationMaster."); }// w w w .j ava 2 s . 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; } }