List of usage examples for org.apache.hadoop.yarn.client.api YarnClientApplication getApplicationSubmissionContext
public ApplicationSubmissionContext getApplicationSubmissionContext()
From source file:de.huberlin.wbi.hiway.common.Client.java
License:Apache License
/** * Main run function for the client.//from ww w . j av a 2 s . c om * * @return true if application completed successfully. */ private boolean run() throws IOException, YarnException { /* log */ System.out.println("Running Client"); yarnClient.start(); YarnClusterMetrics clusterMetrics = yarnClient.getYarnClusterMetrics(); /* log */ System.out.println( "Got Cluster metric info from ASM" + ", numNodeManagers=" + clusterMetrics.getNumNodeManagers()); List<NodeReport> clusterNodeReports = yarnClient.getNodeReports(NodeState.RUNNING); /* log */ System.out.println("Got Cluster node info from ASM"); /* log */ for (NodeReport node : clusterNodeReports) System.out.println("Got node report from ASM for" + ", nodeId=" + node.getNodeId() + ", nodeAddress" + node.getHttpAddress() + ", nodeRackName" + node.getRackName() + ", nodeNumContainers" + node.getNumContainers()); QueueInfo queueInfo = yarnClient.getQueueInfo(this.amQueue); /* log */ System.out.println("Queue info" + ", queueName=" + queueInfo.getQueueName() + ", queueCurrentCapacity=" + queueInfo.getCurrentCapacity() + ", queueMaxCapacity=" + queueInfo.getMaximumCapacity() + ", queueApplicationCount=" + queueInfo.getApplications().size() + ", queueChildQueueCount=" + queueInfo.getChildQueues().size()); List<QueueUserACLInfo> listAclInfo = yarnClient.getQueueAclsInfo(); /* log */ for (QueueUserACLInfo aclInfo : listAclInfo) for (QueueACL userAcl : aclInfo.getUserAcls()) System.out.println("User ACL Info for Queue" + ", queueName=" + aclInfo.getQueueName() + ", userAcl=" + userAcl.name()); // Get a new application id YarnClientApplication app = yarnClient.createApplication(); GetNewApplicationResponse appResponse = app.getNewApplicationResponse(); // Get min/max resource capabilities from RM and change memory ask if needed int maxVC = appResponse.getMaximumResourceCapability().getVirtualCores(); /* log */ System.out.println("Max vCores capabililty of resources in this cluster " + maxVC); int maxMem = appResponse.getMaximumResourceCapability().getMemory(); /* log */ System.out.println("Max mem capabililty of resources in this cluster " + maxMem); // A resource ask cannot exceed the max. if (amVCores > maxVC) { /* log */ System.out.println("AM vCores specified above max threshold of cluster. Using max value." + ", specified=" + amVCores + ", max=" + maxVC); amVCores = maxVC; } if (amMemory > maxMem) { /* log */ System.out.println("AM memory specified above max threshold of cluster. Using max value." + ", specified=" + amMemory + ", max=" + maxMem); amMemory = maxMem; } // set the application name ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); appContext.setApplicationType(conf.get(HiWayConfiguration.HIWAY_AM_APPLICATION_TYPE, HiWayConfiguration.HIWAY_AM_APPLICATION_TYPE_DEFAULT)); appContext.setApplicationName("run " + workflowParam + " (type: " + workflowType.toString() + ")"); ApplicationId appId = appContext.getApplicationId(); String hdfsBaseDirectoryName = conf.get(HiWayConfiguration.HIWAY_AM_DIRECTORY_BASE, HiWayConfiguration.HIWAY_AM_DIRECTORY_BASE_DEFAULT); String hdfsSandboxDirectoryName = conf.get(HiWayConfiguration.HIWAY_AM_DIRECTORY_CACHE, HiWayConfiguration.HIWAY_AM_DIRECTORY_CACHE_DEFAULT); Path hdfsBaseDirectory = new Path(new Path(hdfs.getUri()), hdfsBaseDirectoryName); Data.setHdfsBaseDirectory(hdfsBaseDirectory); Path hdfsSandboxDirectory = new Path(hdfsBaseDirectory, hdfsSandboxDirectoryName); Path hdfsApplicationDirectory = new Path(hdfsSandboxDirectory, appId.toString()); Data.setHdfsApplicationDirectory(hdfsApplicationDirectory); Data.setHdfs(hdfs); Path wfSource, wfDest, wfTemp = null; try { wfSource = new Path(new URI(workflowParam).getPath()); } catch (URISyntaxException e) { wfSource = new Path(workflowParam); } wfDest = new Path(hdfsApplicationDirectory + "/" + wfSource.getName()); // (1) if workflow file in hdfs, then transfer to temp file in local fs if (hdfs.exists(wfSource)) { wfTemp = new Path("./." + wfSource.getName()); System.out.println("Workflow found in HDFS at location " + wfSource); hdfs.copyToLocalFile(false, wfSource, wfTemp); } // (2) if galaxy workflow, then copy and replace input ports if (workflowType.equals(HiWayConfiguration.HIWAY_WORKFLOW_LANGUAGE_OPTS.galaxy)) { wfTemp = preProcessGalaxyWorkflow(wfSource, wfTemp); } if (wfTemp != null) { hdfs.copyFromLocalFile(wfTemp, wfDest); new File(wfTemp.toString()).delete(); } else { hdfs.copyFromLocalFile(wfSource, wfDest); } if (summaryPath != null) summary = new Data(summaryPath); if (customMemPath != null) (new Data(customMemPath)).stageOut(); // Set up the container launch context for the application master ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class); /* set the env variables to be setup in the env where the application master will be run */ System.out.println("Set the environment for the application master"); Map<String, String> env = new HashMap<>(); StringBuilder classPathEnv = new StringBuilder(Environment.CLASSPATH.$()).append(File.pathSeparatorChar) .append("./*"); for (String c : conf.getStrings(YarnConfiguration.YARN_APPLICATION_CLASSPATH, YarnConfiguration.DEFAULT_YARN_APPLICATION_CLASSPATH)) { classPathEnv.append(':'); classPathEnv.append(File.pathSeparatorChar); classPathEnv.append(c.trim()); } if (conf.getBoolean(YarnConfiguration.IS_MINI_YARN_CLUSTER, false)) { classPathEnv.append(':'); classPathEnv.append(System.getProperty("java.class.path")); } env.put("CLASSPATH", classPathEnv.toString()); amContainer.setEnvironment(env); // Set the necessary command to execute the application master Vector<CharSequence> vargs = new Vector<>(30); // Set java executable command System.out.println("Setting up app master command"); vargs.add(Environment.JAVA_HOME.$() + "/bin/java"); if (HiWayConfiguration.debug) vargs.add( "-Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=9010 -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false"); // Set Xmx based on am memory size vargs.add("-Xmx" + amMemory + "m"); vargs.add("-Xss" + "16m"); // Set class name switch (workflowType) { case dax: vargs.add(HiWayConfiguration.HIWAY_WORKFLOW_LANGUAGE_DAX_AM_CLASS); break; case log: vargs.add(HiWayConfiguration.HIWAY_WORKFLOW_LANGUAGE_LOG_AM_CLASS); break; case galaxy: vargs.add(HiWayConfiguration.HIWAY_WORKFLOW_LANGUAGE_GALAXY_AM_CLASS); break; case cuneiformE: vargs.add(HiWayConfiguration.HIWAY_WORKFLOW_LANGUAGE_CUNEIFORME_AM_CLASS); break; default: vargs.add(HiWayConfiguration.HIWAY_WORKFLOW_LANGUAGE_CUNEIFORMJ_AM_CLASS); } vargs.add("--scheduler " + schedulerName.toString()); if (memory != null) vargs.add("--memory " + memory); if (summary != null) vargs.add("--summary " + summary.getName()); if (customMemPath != null) vargs.add("--custom " + customMemPath); vargs.add("--appid " + appId.toString()); if (HiWayConfiguration.debug) vargs.add("--debug"); if (HiWayConfiguration.verbose) vargs.add("--verbose"); vargs.add(workflowParam); vargs.add("> >(tee AppMaster.stdout " + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stdout)"); vargs.add("2> >(tee AppMaster.stderr " + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stderr >&2)"); // Get final command StringBuilder command = new StringBuilder(); for (CharSequence str : vargs) { command.append(str).append(" "); } System.out.println("Completed setting up app master command " + command.toString()); List<String> commands = new ArrayList<>(); commands.add(command.toString()); amContainer.setCommands(commands); // Set up resource type requirements Resource capability = Records.newRecord(Resource.class); capability.setVirtualCores(amVCores); capability.setMemory(amMemory); appContext.setResource(capability); // Setup security tokens if (UserGroupInformation.isSecurityEnabled()) { Credentials credentials = new Credentials(); String tokenRenewer = conf.get(YarnConfiguration.RM_PRINCIPAL); if (tokenRenewer == null || tokenRenewer.length() == 0) { throw new IOException("Can't get Master Kerberos principal for the RM to use as renewer"); } // For now, only getting tokens for the default file-system. final Token<?> tokens[] = hdfs.addDelegationTokens(tokenRenewer, credentials); if (tokens != null) { for (Token<?> token : tokens) { System.out.println("Got dt for " + hdfs.getUri() + "; " + token); } } try (DataOutputBuffer dob = new DataOutputBuffer()) { credentials.writeTokenStorageToStream(dob); ByteBuffer fsTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); amContainer.setTokens(fsTokens); } } appContext.setAMContainerSpec(amContainer); // Set the priority for the application master Priority pri = Records.newRecord(Priority.class); pri.setPriority(amPriority); appContext.setPriority(pri); // Set the queue to which this application is to be submitted in the RM appContext.setQueue(amQueue); // Submit the application to the applications manager /* log */ System.out.println("Submitting application to ASM"); yarnClient.submitApplication(appContext); // Monitor the application boolean success = monitorApplication(appId); if (success && summary != null) { summary.stageIn(); } return success; }
From source file:edu.uci.ics.asterix.aoya.AsterixYARNClient.java
License:Apache License
/** * Initialize and register the application attempt with the YARN ResourceManager. * // www .j a va 2s .c om * @return * @throws IOException * @throws YarnException */ public YarnClientApplication makeApplicationContext() throws IOException, YarnException { //first check to see if an instance already exists. FileSystem fs = FileSystem.get(conf); Path lock = new Path(fs.getHomeDirectory(), CONF_DIR_REL + instanceFolder + instanceLock); LOG.info("Running Deployment"); yarnClient.start(); if (fs.exists(lock)) { ApplicationId lockAppId = getLockFile(); try { ApplicationReport previousAppReport = yarnClient.getApplicationReport(lockAppId); YarnApplicationState prevStatus = previousAppReport.getYarnApplicationState(); if (!(prevStatus == YarnApplicationState.FAILED || prevStatus == YarnApplicationState.KILLED || prevStatus == YarnApplicationState.FINISHED) && mode != Mode.DESTROY && mode != Mode.BACKUP && mode != Mode.RESTORE) { throw new IllegalStateException("Instance is already running in: " + lockAppId); } else if (mode != Mode.DESTROY && mode != Mode.BACKUP && mode != Mode.RESTORE) { //stale lock file LOG.warn("Stale lockfile detected. Instance attempt " + lockAppId + " may have exited abnormally"); deleteLockFile(); } } catch (YarnException e) { LOG.warn( "Stale lockfile detected, but the RM has no record of this application's last run. This is normal if the cluster was restarted."); deleteLockFile(); } } // Get a new application id YarnClientApplication app = yarnClient.createApplication(); GetNewApplicationResponse appResponse = app.getNewApplicationResponse(); int maxMem = appResponse.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); // A resource ask cannot exceed the max. if (amMemory > maxMem) { LOG.info("AM memory specified above max threshold of cluster. Using max value." + ", specified=" + amMemory + ", max=" + maxMem); amMemory = maxMem; } // set the application name ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); appContext.setApplicationName(appName); return app; }
From source file:edu.uci.ics.asterix.aoya.AsterixYARNClient.java
License:Apache License
/** * Submits the request to start the AsterixApplicationMaster to the YARN ResourceManager. * /*from w w w. ja v a 2 s . c o m*/ * @param app * The application attempt handle. * @param resources * Resources to be distributed as part of the container launch * @param mode * The mode of the ApplicationMaster * @return The application ID of the new Asterix instance. * @throws IOException * @throws YarnException */ public ApplicationId deployAM(YarnClientApplication app, List<DFSResourceCoordinate> resources, Mode mode) throws IOException, YarnException { // Set up the container launch context for the application master ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class); ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); // Set local resource info into app master container launch context Map<String, LocalResource> localResources = new HashMap<String, LocalResource>(); for (DFSResourceCoordinate res : resources) { localResources.put(res.name, res.res); } amContainer.setLocalResources(localResources); // Set the env variables to be setup in the env where the application // master will be run LOG.info("Set the environment for the application master"); Map<String, String> env = new HashMap<String, String>(); // using the env info, the application master will create the correct // local resource for the // eventual containers that will be launched to execute the shell // scripts for (DFSResourceCoordinate res : resources) { if (res.envs == null) { //some entries may not have environment variables. continue; } for (Map.Entry<String, String> e : res.envs.entrySet()) { env.put(e.getValue(), e.getKey()); } } //this is needed for when the RM address isn't known from the environment of the AM env.put(AConstants.RMADDRESS, conf.get("yarn.resourcemanager.address")); env.put(AConstants.RMSCHEDULERADDRESS, conf.get("yarn.resourcemanager.scheduler.address")); ///add miscellaneous environment variables. env.put(AConstants.INSTANCESTORE, CONF_DIR_REL + instanceFolder); env.put(AConstants.DFS_BASE, FileSystem.get(conf).getHomeDirectory().toUri().toString()); env.put(AConstants.CC_JAVA_OPTS, ccJavaOpts); env.put(AConstants.NC_JAVA_OPTS, ncJavaOpts); // Add AppMaster.jar location to classpath // At some point we should not be required to add // the hadoop specific classpaths to the env. // It should be provided out of the box. // For now setting all required classpaths including // the classpath to "." for the application jar StringBuilder classPathEnv = new StringBuilder("").append("./*"); for (String c : conf.getStrings(YarnConfiguration.YARN_APPLICATION_CLASSPATH, YarnConfiguration.DEFAULT_YARN_APPLICATION_CLASSPATH)) { classPathEnv.append(File.pathSeparatorChar); classPathEnv.append(c.trim()); } classPathEnv.append(File.pathSeparatorChar).append("." + File.separator + "log4j.properties"); // add the runtime classpath needed for tests to work if (conf.getBoolean(YarnConfiguration.IS_MINI_YARN_CLUSTER, false)) { LOG.info("In YARN MiniCluster"); classPathEnv.append(System.getProperty("path.separator")); classPathEnv.append(System.getProperty("java.class.path")); env.put("HADOOP_CONF_DIR", System.getProperty("user.dir") + File.separator + "target" + File.separator); } LOG.info("AM Classpath:" + classPathEnv.toString()); env.put("CLASSPATH", classPathEnv.toString()); amContainer.setEnvironment(env); // Set the necessary command to execute the application master Vector<CharSequence> vargs = new Vector<CharSequence>(30); // Set java executable command LOG.info("Setting up app master command"); vargs.add(JAVA_HOME + File.separator + "bin" + File.separator + "java"); // Set class name vargs.add(appMasterMainClass); //Set params for Application Master if (debugFlag) { vargs.add("-debug"); } if (mode == Mode.DESTROY) { vargs.add("-obliterate"); } else if (mode == Mode.BACKUP) { vargs.add("-backup"); } else if (mode == Mode.RESTORE) { vargs.add("-restore " + snapName); } else if (mode == Mode.INSTALL) { vargs.add("-initial "); } if (refresh) { vargs.add("-refresh"); } //vargs.add("/bin/ls -alh asterix-server.zip/repo"); vargs.add("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + File.separator + "AppMaster.stdout"); vargs.add("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + File.separator + "AppMaster.stderr"); // Get final commmand StringBuilder command = new StringBuilder(); for (CharSequence str : vargs) { command.append(str).append(" "); } LOG.info("Completed setting up app master command " + command.toString()); List<String> commands = new ArrayList<String>(); commands.add(command.toString()); amContainer.setCommands(commands); // Set up resource type requirements // For now, only memory is supported so we set memory requirements Resource capability = Records.newRecord(Resource.class); capability.setMemory(amMemory); appContext.setResource(capability); // Service data is a binary blob that can be passed to the application // Not needed in this scenario // amContainer.setServiceData(serviceData); // The following are not required for launching an application master // amContainer.setContainerId(containerId); appContext.setAMContainerSpec(amContainer); // Set the priority for the application master Priority pri = Records.newRecord(Priority.class); // TODO - what is the range for priority? how to decide? pri.setPriority(amPriority); appContext.setPriority(pri); // Set the queue to which this application is to be submitted in the RM appContext.setQueue(amQueue); // Submit the application to the applications manager // SubmitApplicationResponse submitResp = // applicationsManager.submitApplication(appRequest); // Ignore the response as either a valid response object is returned on // success // or an exception thrown to denote some form of a failure LOG.info("Submitting application to ASM"); yarnClient.submitApplication(appContext); //now write the instance lock if (mode == Mode.INSTALL || mode == Mode.START) { FileSystem fs = FileSystem.get(conf); Path lock = new Path(fs.getHomeDirectory(), CONF_DIR_REL + instanceFolder + instanceLock); if (fs.exists(lock)) { throw new IllegalStateException("Somehow, this instance has been launched twice. "); } BufferedWriter out = new BufferedWriter(new OutputStreamWriter(fs.create(lock, true))); try { out.write(app.getApplicationSubmissionContext().getApplicationId().toString()); out.close(); } finally { out.close(); } } return app.getApplicationSubmissionContext().getApplicationId(); }
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 ww . ja va2 s. c o m*/ //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
/** * Setup and submit the Gobblin Yarn application. * * @throws IOException if there's anything wrong setting up and submitting the Yarn application * @throws YarnException if there's anything wrong setting up and submitting the Yarn application *///from www . j a v a 2 s . c o m @VisibleForTesting ApplicationId setupAndSubmitApplication() throws IOException, YarnException { YarnClientApplication gobblinYarnApp = this.yarnClient.createApplication(); ApplicationSubmissionContext appSubmissionContext = gobblinYarnApp.getApplicationSubmissionContext(); appSubmissionContext.setApplicationType(GOBBLIN_YARN_APPLICATION_TYPE); ApplicationId applicationId = appSubmissionContext.getApplicationId(); GetNewApplicationResponse newApplicationResponse = gobblinYarnApp.getNewApplicationResponse(); // Set up resource type requirements for ApplicationMaster Resource resource = prepareContainerResource(newApplicationResponse); // Add lib jars, and jars and files that the ApplicationMaster need as LocalResources Map<String, LocalResource> appMasterLocalResources = addAppMasterLocalResources(applicationId); ContainerLaunchContext amContainerLaunchContext = Records.newRecord(ContainerLaunchContext.class); amContainerLaunchContext.setLocalResources(appMasterLocalResources); amContainerLaunchContext.setEnvironment(YarnHelixUtils.getEnvironmentVariables(this.yarnConfiguration)); amContainerLaunchContext .setCommands(Lists.newArrayList(buildApplicationMasterCommand(resource.getMemory()))); if (UserGroupInformation.isSecurityEnabled()) { setupSecurityTokens(amContainerLaunchContext); } // Setup the application submission context appSubmissionContext.setApplicationName(this.applicationName); appSubmissionContext.setResource(resource); appSubmissionContext.setQueue(this.appQueueName); appSubmissionContext.setPriority(Priority.newInstance(0)); appSubmissionContext.setAMContainerSpec(amContainerLaunchContext); // Also setup container local resources by copying local jars and files the container need to HDFS addContainerLocalResources(applicationId); // Submit the application LOGGER.info("Submitting application " + applicationId); this.yarnClient.submitApplication(appSubmissionContext); LOGGER.info("Application successfully submitted and accepted"); ApplicationReport applicationReport = this.yarnClient.getApplicationReport(applicationId); LOGGER.info("Application Name: " + applicationReport.getName()); LOGGER.info("Application Tracking URL: " + applicationReport.getTrackingUrl()); LOGGER.info("Application User: " + applicationReport.getUser() + " Queue: " + applicationReport.getQueue()); return applicationId; }
From source file:hadoop.yarn.distributedshell.DshellClient.java
License:Apache License
/** * Main run function for the client//w ww . ja v a 2 s.c o m * * @return true if application completed successfully * @throws IOException * @throws YarnException */ public boolean run() throws IOException, YarnException { LOG.info("Running Client"); yarnClient.start(); YarnClusterMetrics clusterMetrics = yarnClient.getYarnClusterMetrics(); LOG.info("Got Cluster metric info from ASM" + ", numNodeManagers=" + clusterMetrics.getNumNodeManagers()); List<NodeReport> clusterNodeReports = yarnClient.getNodeReports(NodeState.RUNNING); LOG.info("Got Cluster node info from ASM"); for (NodeReport node : clusterNodeReports) { LOG.info("Got node report from ASM for" + ", nodeId=" + node.getNodeId() + ", nodeAddress" + node.getHttpAddress() + ", nodeRackName" + node.getRackName() + ", nodeNumContainers" + node.getNumContainers()); } QueueInfo queueInfo = yarnClient.getQueueInfo(this.amQueue); LOG.info("Queue info" + ", queueName=" + queueInfo.getQueueName() + ", queueCurrentCapacity=" + queueInfo.getCurrentCapacity() + ", queueMaxCapacity=" + queueInfo.getMaximumCapacity() + ", queueApplicationCount=" + queueInfo.getApplications().size() + ", queueChildQueueCount=" + queueInfo.getChildQueues().size()); List<QueueUserACLInfo> listAclInfo = yarnClient.getQueueAclsInfo(); for (QueueUserACLInfo aclInfo : listAclInfo) { for (QueueACL userAcl : aclInfo.getUserAcls()) { LOG.info("User ACL Info for Queue" + ", queueName=" + aclInfo.getQueueName() + ", userAcl=" + userAcl.name()); } } // Get a new application id YarnClientApplication app = yarnClient.createApplication(); GetNewApplicationResponse appResponse = app.getNewApplicationResponse(); // TODO get min/max resource capabilities from RM and change memory ask // if needed // If we do not have min/max, we may not be able to correctly request // the required resources from the RM for the app master // Memory ask has to be a multiple of min and less than max. // Dump out information about cluster capability as seen by the resource // manager int maxMem = appResponse.getMaximumResourceCapability().getMemory(); LOG.info("Max mem capabililty of resources in this cluster " + maxMem); // A resource ask cannot exceed the max. if (amMemory > maxMem) { LOG.info("AM memory specified above max threshold of cluster. Using max value." + ", specified=" + amMemory + ", max=" + maxMem); amMemory = maxMem; } int maxVCores = appResponse.getMaximumResourceCapability().getVirtualCores(); LOG.info("Max virtual cores capabililty of resources in this cluster " + maxVCores); if (amVCores > maxVCores) { LOG.info("AM virtual cores specified above max threshold of cluster. " + "Using max value." + ", specified=" + amVCores + ", max=" + maxVCores); amVCores = maxVCores; } // set the application name ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); ApplicationId appId = appContext.getApplicationId(); appContext.setKeepContainersAcrossApplicationAttempts(keepContainers); appContext.setApplicationName(appName); // Set up the container launch context for the application master ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class); // set local resources for the application master // local files or archives as needed // In this scenario, the jar file for the application master is part of // the local resources Map<String, LocalResource> localResources = new HashMap<String, LocalResource>(); 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 FileSystem fs = FileSystem.get(conf); addToLocalResources(fs, appMasterJar, appMasterJarPath, appId.toString(), localResources, null); // Set the log4j properties if needed if (!log4jPropFile.isEmpty()) { addToLocalResources(fs, log4jPropFile, log4jPath, appId.toString(), localResources, null); } // The shell script has to be made available on the final container(s) // where it will be executed. // To do this, we need to first copy into the filesystem that is visible // to the yarn framework. // We do not need to set this as a local resource for the application // master as the application master does not need it. String hdfsShellScriptLocation = ""; long hdfsShellScriptLen = 0; long hdfsShellScriptTimestamp = 0; if (!shellScriptPath.isEmpty()) { Path shellSrc = new Path(shellScriptPath); String shellPathSuffix = appName + "/" + appId.toString() + "/" + SCRIPT_PATH; Path shellDst = new Path(fs.getHomeDirectory(), shellPathSuffix); fs.copyFromLocalFile(false, true, shellSrc, shellDst); hdfsShellScriptLocation = shellDst.toUri().toString(); FileStatus shellFileStatus = fs.getFileStatus(shellDst); hdfsShellScriptLen = shellFileStatus.getLen(); hdfsShellScriptTimestamp = shellFileStatus.getModificationTime(); } if (!shellCommand.isEmpty()) { addToLocalResources(fs, null, shellCommandPath, appId.toString(), localResources, shellCommand); } if (shellArgs.length > 0) { addToLocalResources(fs, null, shellArgsPath, appId.toString(), localResources, StringUtils.join(shellArgs, " ")); } // Set local resource info into app master container launch context amContainer.setLocalResources(localResources); // Set the necessary security tokens as needed // amContainer.setContainerTokens(containerToken); // Set the env variables to be setup in the env where the application // master will be run LOG.info("Set the environment for the application master"); Map<String, String> env = new HashMap<String, String>(); // put location of shell script into env // using the env info, the application master will create the correct // local resource for the // eventual containers that will be launched to execute the shell // scripts env.put(DshellDSConstants.DISTRIBUTEDSHELLSCRIPTLOCATION, hdfsShellScriptLocation); env.put(DshellDSConstants.DISTRIBUTEDSHELLSCRIPTTIMESTAMP, Long.toString(hdfsShellScriptTimestamp)); env.put(DshellDSConstants.DISTRIBUTEDSHELLSCRIPTLEN, Long.toString(hdfsShellScriptLen)); // ========================================jar? if (containerJarPaths.length != 0) { for (int i = 0; i < containerJarPaths.length; i++) { String hdfsJarLocation = ""; String[] jarNameSplit = containerJarPaths[i].split("/"); String jarName = jarNameSplit[jarNameSplit.length - 1]; long hdfsJarLen = 0; long hdfsJarTimestamp = 0; if (!containerJarPaths[i].isEmpty()) { Path jarSrc = new Path(containerJarPaths[i]); String jarPathSuffix = appName + "/" + appId.toString() + "/" + jarName; Path jarDst = new Path(fs.getHomeDirectory(), jarPathSuffix); fs.copyFromLocalFile(false, true, jarSrc, jarDst); hdfsJarLocation = jarDst.toUri().toString(); FileStatus jarFileStatus = fs.getFileStatus(jarDst); hdfsJarLen = jarFileStatus.getLen(); hdfsJarTimestamp = jarFileStatus.getModificationTime(); env.put(DshellDSConstants.DISTRIBUTEDJARLOCATION + i, hdfsJarLocation); env.put(DshellDSConstants.DISTRIBUTEDJARTIMESTAMP + i, Long.toString(hdfsJarTimestamp)); env.put(DshellDSConstants.DISTRIBUTEDJARLEN + i, Long.toString(hdfsJarLen)); } } } // ========================================jar? // ========================================archive? if (containerArchivePaths.length != 0) { for (int i = 0; i < containerArchivePaths.length; i++) { String hdfsArchiveLocation = ""; String[] archiveNameSplit = containerArchivePaths[i].split("/"); String archiveName = archiveNameSplit[archiveNameSplit.length - 1]; long hdfsArchiveLen = 0; long hdfsArchiveTimestamp = 0; if (!containerArchivePaths[i].isEmpty()) { Path archiveSrc = new Path(containerArchivePaths[i]); String archivePathSuffix = appName + "/" + appId.toString() + "/" + archiveName; Path archiveDst = new Path(fs.getHomeDirectory(), archivePathSuffix); fs.copyFromLocalFile(false, true, archiveSrc, archiveDst); hdfsArchiveLocation = archiveDst.toUri().toString(); FileStatus archiveFileStatus = fs.getFileStatus(archiveDst); hdfsArchiveLen = archiveFileStatus.getLen(); hdfsArchiveTimestamp = archiveFileStatus.getModificationTime(); env.put(DshellDSConstants.DISTRIBUTEDARCHIVELOCATION + i, hdfsArchiveLocation); env.put(DshellDSConstants.DISTRIBUTEDARCHIVETIMESTAMP + i, Long.toString(hdfsArchiveTimestamp)); env.put(DshellDSConstants.DISTRIBUTEDARCHIVELEN + i, Long.toString(hdfsArchiveLen)); } } } // ========================================archive? // Add AppMaster.jar location to classpath // At some point we should not be required to add // the hadoop specific classpaths to the env. // It should be provided out of the box. // For now setting all required classpaths including // the classpath to "." for the application jar StringBuilder classPathEnv = new StringBuilder(Environment.CLASSPATH.$$()) .append(ApplicationConstants.CLASS_PATH_SEPARATOR).append("./*"); for (String c : conf.getStrings(YarnConfiguration.YARN_APPLICATION_CLASSPATH, YarnConfiguration.DEFAULT_YARN_CROSS_PLATFORM_APPLICATION_CLASSPATH)) { classPathEnv.append(ApplicationConstants.CLASS_PATH_SEPARATOR); classPathEnv.append(c.trim()); } classPathEnv.append(ApplicationConstants.CLASS_PATH_SEPARATOR).append("./log4j.properties"); // add the runtime classpath needed for tests to work if (conf.getBoolean(YarnConfiguration.IS_MINI_YARN_CLUSTER, false)) { classPathEnv.append(':'); classPathEnv.append(System.getProperty("java.class.path")); } env.put("CLASSPATH", classPathEnv.toString()); amContainer.setEnvironment(env); // Set the necessary command to execute the application master Vector<CharSequence> vargs = new Vector<CharSequence>(30); // Set java executable command LOG.info("Setting up app master command"); vargs.add(Environment.JAVA_HOME.$$() + "/bin/java"); // Set Xmx based on am memory size vargs.add("-Xmx" + amMemory + "m"); // Set class name vargs.add(appMasterMainClass); // Set params for Application Master vargs.add("--container_memory " + String.valueOf(containerMemory)); vargs.add("--container_vcores " + String.valueOf(containerVirtualCores)); vargs.add("--num_containers " + String.valueOf(numContainers)); vargs.add("--priority " + String.valueOf(shellCmdPriority)); for (Map.Entry<String, String> entry : shellEnv.entrySet()) { vargs.add("--shell_env " + entry.getKey() + "=" + entry.getValue()); } if (debugFlag) { vargs.add("--debug"); } vargs.add("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stdout"); vargs.add("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/AppMaster.stderr"); // Get final commmand StringBuilder command = new StringBuilder(); for (CharSequence str : vargs) { command.append(str).append(" "); } LOG.info("Completed setting up app master command " + command.toString()); List<String> commands = new ArrayList<String>(); commands.add(command.toString()); amContainer.setCommands(commands); // Set up resource type requirements // For now, both memory and vcores are supported, so we set memory and // vcores requirements Resource capability = Records.newRecord(Resource.class); capability.setMemory(amMemory); capability.setVirtualCores(amVCores); appContext.setResource(capability); // Service data is a binary blob that can be passed to the application // Not needed in this scenario // amContainer.setServiceData(serviceData); // Setup security tokens if (UserGroupInformation.isSecurityEnabled()) { Credentials credentials = new Credentials(); String tokenRenewer = conf.get(YarnConfiguration.RM_PRINCIPAL); if (tokenRenewer == null || tokenRenewer.length() == 0) { throw new IOException("Can't get Master Kerberos principal for the RM to use as renewer"); } // For now, only getting tokens for the default file-system. final Token<?> tokens[] = fs.addDelegationTokens(tokenRenewer, credentials); if (tokens != null) { for (Token<?> token : tokens) { LOG.info("Got dt for " + fs.getUri() + "; " + token); } } DataOutputBuffer dob = new DataOutputBuffer(); credentials.writeTokenStorageToStream(dob); ByteBuffer fsTokens = ByteBuffer.wrap(dob.getData(), 0, dob.getLength()); amContainer.setTokens(fsTokens); } appContext.setAMContainerSpec(amContainer); // Set the priority for the application master Priority pri = Records.newRecord(Priority.class); // TODO - what is the range for priority? how to decide? pri.setPriority(amPriority); appContext.setPriority(pri); // Set the queue to which this application is to be submitted in the RM appContext.setQueue(amQueue); // Submit the application to the applications manager // SubmitApplicationResponse submitResp = // applicationsManager.submitApplication(appRequest); // Ignore the response as either a valid response object is returned on // success // or an exception thrown to denote some form of a failure LOG.info("Submitting application to ASM"); yarnClient.submitApplication(appContext); // TODO // Try submitting the same request again // app submission failure? // Monitor the application return monitorApplication(appId); }
From source file:husky.client.HuskyYarnClient.java
License:Apache License
private boolean run() throws YarnException, IOException { mYarnClient.start();/* w w w . ja va 2 s.c o m*/ YarnClientApplication app = mYarnClient.createApplication(); ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); appContext.setApplicationName(mAppName); Resource resource = Records.newRecord(Resource.class); resource.setMemory(mAppMasterMemory); resource.setVirtualCores(mNumVirtualCores); appContext.setResource(resource); mAppId = appContext.getApplicationId(); ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class); amContainer.setLocalResources(getLocalResources()); amContainer.setEnvironment(getEnvironment()); StringBuilder cmdBuilder = new StringBuilder(); if (!mKeyTabFile.isEmpty()) { if (mUserName.isEmpty()) { throw new RuntimeException("Username is not given but is required by kinit"); } cmdBuilder.append("kinit -kt ").append(mKeyTabFile).append(" -V ").append(mUserName).append(" && "); } cmdBuilder.append(JAVA_HOME.$()).append("/bin/java").append(" -Xmx").append(mAppMasterMemory).append("m ") .append(HuskyApplicationMaster.class.getName()).append(" --container_memory ") .append(mContainerMemory).append(" --container_vcores ").append(mNumVirtualCores) .append(" --app_priority ").append(mAppPriority).append(" --app_master_log_dir <LOG_DIR>") .append(" --master ").append(mMasterExec).append(" --application ").append(mAppExec) .append(" --config ").append(mConfigFile); if (!mLdLibraryPath.isEmpty()) { cmdBuilder.append(" --ld_library_path \"").append(mLdLibraryPath).append("\" "); } cmdBuilder.append(" --worker_infos ").append(mWorkerInfos.get(0).getFirst()).append(":") .append(mWorkerInfos.get(0).getSecond()); for (int i = 1; i < mWorkerInfos.size(); i++) { cmdBuilder.append(',').append(mWorkerInfos.get(i).getFirst()).append(":") .append(mWorkerInfos.get(i).getSecond()); } if (!mLocalFiles.isEmpty()) { cmdBuilder.append(" --local_files \"").append(mLocalFiles).append("\""); } if (!mLocalArchives.isEmpty()) { cmdBuilder.append(" --local_archives \"").append(mLocalArchives).append("\""); } if (!mLogPathToHDFS.isEmpty()) { cmdBuilder.append(" --log_to_hdfs \"").append(mLogPathToHDFS).append("\""); } cmdBuilder.append(" 1>").append("<LOG_DIR>/HuskyAppMaster.stdout").append(" 2>") .append("<LOG_DIR>/HuskyAppMaster.stderr"); if (!mLogPathToHDFS.isEmpty()) { cmdBuilder.append("; am_exit_code=$?").append("; hadoop fs -put -f <LOG_DIR>/HuskyAppMaster.stdout ") .append(mLogPathToHDFS).append("; hadoop fs -put -f <LOG_DIR>/HuskyAppMaster.stderr ") .append(mLogPathToHDFS).append("; exit \"$am_exit_code\""); } amContainer.setCommands(Collections.singletonList(cmdBuilder.toString())); LOG.info("Command: " + amContainer.getCommands().get(0)); appContext.setAMContainerSpec(amContainer); mYarnClient.submitApplication(appContext); return monitorApp(); }
From source file:hws.core.JobClient.java
License:Apache License
public void run(String[] args) throws Exception { //final String command = args[0]; //final int n = Integer.valueOf(args[1]); //final Path jarPath = new Path(args[2]); Options options = new Options(); /*options.addOption(OptionBuilder.withLongOpt("jar") .withDescription( "Jar path" ) .hasArg()//from w w w . ja v a 2 s. co m .withArgName("JarPath") .create()); options.addOption(OptionBuilder.withLongOpt("scheduler") .withDescription( "Scheduler class name" ) .hasArg() .withArgName("ClassName") .create()); */options.addOption(OptionBuilder.withLongOpt("zk-servers") .withDescription("List of the ZooKeeper servers").hasArgs().withArgName("zkAddrs").create("zks")); //options.addOption("l", "list", false, "list modules"); options.addOption(OptionBuilder.withLongOpt("load").withDescription("load new modules").hasArgs() .withArgName("XMLFiles").create()); /*options.addOption(OptionBuilder.withLongOpt( "remove" ) .withDescription( "remove modules" ) .hasArgs() .withArgName("ModuleNames") .create("rm")); */CommandLineParser parser = new BasicParser(); CommandLine cmd = parser.parse(options, args); //Path jarPath = null; //String schedulerClassName = null; String[] xmlFileNames = null; //String []moduleNames = null; String zksArgs = ""; String[] zkServers = null; if (cmd.hasOption("zks")) { zksArgs = "-zks"; zkServers = cmd.getOptionValues("zks"); for (String zks : zkServers) { zksArgs += " " + zks; } } //Logger setup //FSDataOutputStream writer = FileSystem.get(conf).create(new Path("hdfs:///hws/apps/"+appIdStr+"/logs/jobClient.log")); //Logger.addOutputStream(writer); /*if(cmd.hasOption("l")){ LOG.warn("Argument --list (-l) is not supported yet."); } if(cmd.hasOption("jar")){ jarPath = new Path(cmd.getOptionValue("jar")); } if(cmd.hasOption("scheduler")){ schedulerClassName = cmd.getOptionValue("scheduler"); }*/ if (cmd.hasOption("load")) { xmlFileNames = cmd.getOptionValues("load"); } /*else if(cmd.hasOption("rm")){ moduleNames = cmd.getOptionValues("rm"); }*/ //LOG.info("Jar-Path "+jarPath); if (xmlFileNames != null) { String paths = ""; for (String path : xmlFileNames) { paths += path + "; "; } LOG.info("Load XMLs: " + paths); } /*if(moduleNames!=null){ String modules = ""; for(String module: moduleNames){ modules += module+"; "; } LOG.info("remove: "+modules); }*/ // Create yarnClient YarnConfiguration conf = new YarnConfiguration(); YarnClient yarnClient = YarnClient.createYarnClient(); yarnClient.init(conf); yarnClient.start(); // Create application via yarnClient YarnClientApplication app = yarnClient.createApplication(); System.out.println("LOG Path: " + ApplicationConstants.LOG_DIR_EXPANSION_VAR); // Set up the container launch context for the application master ContainerLaunchContext amContainer = Records.newRecord(ContainerLaunchContext.class); ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); ApplicationId appId = appContext.getApplicationId(); ZkClient zk = new ZkClient(zkServers[0]); //TODO select a ZooKeeper server if (!zk.exists("/hadoop-watershed")) { zk.createPersistent("/hadoop-watershed", ""); } zk.createPersistent("/hadoop-watershed/" + appId.toString(), ""); FileSystem fs = FileSystem.get(conf); LOG.info("Collecting files to upload"); fs.mkdirs(new Path("hdfs:///hws/apps/" + appId.toString())); fs.mkdirs(new Path("hdfs:///hws/apps/" + appId.toString() + "/logs")); ModulePipeline modulePipeline = ModulePipeline.fromXMLFiles(xmlFileNames); LOG.info("Uploading files to HDFS"); for (String path : modulePipeline.files()) { uploadFile(fs, new File(path), appId); } LOG.info("Upload finished"); String modulePipelineJson = Json.dumps(modulePipeline); String modulePipelineBase64 = Base64.encodeBase64String(StringUtils.getBytesUtf8(modulePipelineJson)) .replaceAll("\\s", ""); LOG.info("ModulePipeline: " + modulePipelineJson); //LOG.info("ModulePipeline: "+modulePipelineBase64); amContainer.setCommands(Collections.singletonList("$JAVA_HOME/bin/java" + " -Xmx256M" + " hws.core.JobMaster" + " -aid " + appId.toString() + " --load " + modulePipelineBase64 + " " + zksArgs + " 1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stdout" + " 2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stderr")); // Setup jar for ApplicationMaster //LocalResource appMasterJar = Records.newRecord(LocalResource.class); //setupAppMasterJar(jarPath, appMasterJar); //amContainer.setLocalResources(Collections.singletonMap("hws.jar", appMasterJar)); LOG.info("Listing files for YARN-Watershed"); RemoteIterator<LocatedFileStatus> filesIterator = fs.listFiles(new Path("hdfs:///hws/bin/"), false); Map<String, LocalResource> resources = new HashMap<String, LocalResource>(); LOG.info("Files setup as resource"); while (filesIterator.hasNext()) { LocatedFileStatus fileStatus = filesIterator.next(); // Setup jar for ApplicationMaster LocalResource containerJar = Records.newRecord(LocalResource.class); ContainerUtils.setupContainerJar(fs, fileStatus.getPath(), containerJar); resources.put(fileStatus.getPath().getName(), containerJar); } LOG.info("container resource setup"); amContainer.setLocalResources(resources); fs.close(); //closing FileSystem interface // Setup CLASSPATH for ApplicationMaster Map<String, String> appMasterEnv = new HashMap<String, String>(); ContainerUtils.setupContainerEnv(appMasterEnv, conf); amContainer.setEnvironment(appMasterEnv); // Set up resource type requirements for ApplicationMaster Resource capability = Records.newRecord(Resource.class); capability.setMemory(256); capability.setVirtualCores(1); // Finally, set-up ApplicationSubmissionContext for the application //ApplicationSubmissionContext appContext = //app.getApplicationSubmissionContext(); appContext.setApplicationName("Hadoop-Watershed"); // application name appContext.setAMContainerSpec(amContainer); appContext.setResource(capability); appContext.setQueue("default"); // queue // Submit application LOG.info("Submitting application " + appId); yarnClient.submitApplication(appContext); LOG.info("Waiting for containers to finish"); zk.waitUntilExists("/hadoop-watershed/" + appId.toString() + "/done", TimeUnit.MILLISECONDS, 250); ApplicationReport appReport = yarnClient.getApplicationReport(appId); YarnApplicationState appState = appReport.getYarnApplicationState(); while (appState != YarnApplicationState.FINISHED && appState != YarnApplicationState.KILLED && appState != YarnApplicationState.FAILED) { Thread.sleep(100); appReport = yarnClient.getApplicationReport(appId); appState = appReport.getYarnApplicationState(); } System.out.println("Application " + appId + " finished with" + " state " + appState + " at " + appReport.getFinishTime()); System.out.println("deleting " + appId.toString() + " znode"); zk.deleteRecursive("/hadoop-watershed/" + appId.toString()); //TODO remove app folder from ZooKeeper }
From source file:io.amient.yarn1.YarnClient.java
License:Open Source License
/** * This method should be called by the implementing application static main * method. It does all the work around creating a yarn application and * submitting the request to the yarn resource manager. The class given in * the appClass argument will be run inside the yarn-allocated master * container./* w w w . j ava 2 s . c o m*/ */ public static void submitApplicationMaster(Properties appConfig, Class<? extends YarnMaster> masterClass, String[] args, Boolean awaitCompletion) throws Exception { log.info("Yarn1 App Configuration:"); for (Object param : appConfig.keySet()) { log.info(param.toString() + " = " + appConfig.get(param).toString()); } String yarnConfigPath = appConfig.getProperty("yarn1.site", "/etc/hadoop"); String masterClassName = masterClass.getName(); appConfig.setProperty("yarn1.master.class", masterClassName); String applicationName = appConfig.getProperty("yarn1.application.name", masterClassName); log.info("--------------------------------------------------------------"); if (Boolean.valueOf(appConfig.getProperty("yarn1.local.mode", "false"))) { YarnMaster.run(appConfig, args); return; } int masterPriority = Integer.valueOf( appConfig.getProperty("yarn1.master.priority", String.valueOf(YarnMaster.DEFAULT_MASTER_PRIORITY))); int masterMemoryMb = Integer.valueOf(appConfig.getProperty("yarn1.master.memory.mb", String.valueOf(YarnMaster.DEFAULT_MASTER_MEMORY_MB))); int masterNumCores = Integer.valueOf( appConfig.getProperty("yarn1.master.num.cores", String.valueOf(YarnMaster.DEFAULT_MASTER_CORES))); String queue = appConfig.getProperty("yarn1.queue"); Configuration yarnConfig = new YarnConfiguration(); yarnConfig.addResource(new FileInputStream(yarnConfigPath + "/core-site.xml")); yarnConfig.addResource(new FileInputStream(yarnConfigPath + "/hdfs-site.xml")); yarnConfig.addResource(new FileInputStream(yarnConfigPath + "/yarn-site.xml")); for (Map.Entry<Object, Object> entry : appConfig.entrySet()) { yarnConfig.set(entry.getKey().toString(), entry.getValue().toString()); } final org.apache.hadoop.yarn.client.api.YarnClient yarnClient = org.apache.hadoop.yarn.client.api.YarnClient .createYarnClient(); yarnClient.init(yarnConfig); yarnClient.start(); for (NodeReport report : yarnClient.getNodeReports(NodeState.RUNNING)) { log.debug("Node report:" + report.getNodeId() + " @ " + report.getHttpAddress() + " | " + report.getCapability()); } log.info("Submitting application master class " + masterClassName); YarnClientApplication app = yarnClient.createApplication(); GetNewApplicationResponse appResponse = app.getNewApplicationResponse(); final ApplicationId appId = appResponse.getApplicationId(); if (appId == null) { System.exit(111); } else { appConfig.setProperty("am.timestamp", String.valueOf(appId.getClusterTimestamp())); appConfig.setProperty("am.id", String.valueOf(appId.getId())); } YarnClient.distributeResources(yarnConfig, appConfig, applicationName); String masterJvmArgs = appConfig.getProperty("yarn1.master.jvm.args", ""); YarnContainerContext masterContainer = new YarnContainerContext(yarnConfig, appConfig, masterJvmArgs, masterPriority, masterMemoryMb, masterNumCores, applicationName, YarnMaster.class, args); ApplicationSubmissionContext appContext = app.getApplicationSubmissionContext(); appContext.setApplicationName(masterClassName); appContext.setResource(masterContainer.capability); appContext.setPriority(masterContainer.priority); appContext.setQueue(queue); appContext.setApplicationType(appConfig.getProperty("yarn1.application.type", "YARN")); appContext.setAMContainerSpec(masterContainer.createContainerLaunchContext()); log.info("Master container spec: " + masterContainer.capability); yarnClient.submitApplication(appContext); ApplicationReport report = yarnClient.getApplicationReport(appId); log.info("Tracking URL: " + report.getTrackingUrl()); if (awaitCompletion) { Runtime.getRuntime().addShutdownHook(new Thread() { @Override public void run() { if (!yarnClient.isInState(Service.STATE.STOPPED)) { log.info("Killing yarn application in shutdown hook"); try { yarnClient.killApplication(appId); } catch (Throwable e) { log.error("Failed to kill yarn application - please check YARN Resource Manager", e); } } } }); float lastProgress = -0.0f; while (true) { try { Thread.sleep(10000); report = yarnClient.getApplicationReport(appId); if (lastProgress != report.getProgress()) { lastProgress = report.getProgress(); log.info(report.getApplicationId() + " " + (report.getProgress() * 100.00) + "% " + (System.currentTimeMillis() - report.getStartTime()) + "(ms) " + report.getDiagnostics()); } if (!report.getFinalApplicationStatus().equals(FinalApplicationStatus.UNDEFINED)) { log.info(report.getApplicationId() + " " + report.getFinalApplicationStatus()); log.info("Tracking url: " + report.getTrackingUrl()); log.info("Finish time: " + ((System.currentTimeMillis() - report.getStartTime()) / 1000) + "(s)"); break; } } catch (Throwable e) { log.error("Master Heart Beat Error - terminating", e); yarnClient.killApplication(appId); Thread.sleep(2000); } } yarnClient.stop(); if (!report.getFinalApplicationStatus().equals(FinalApplicationStatus.SUCCEEDED)) { System.exit(112); } } yarnClient.stop(); }
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./*from w w w . ja va 2 s . c om*/ */ 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); }