List of usage examples for org.apache.hadoop.yarn.api.records ApplicationReport getYarnApplicationState
@Public @Stable public abstract YarnApplicationState getYarnApplicationState();
YarnApplicationState of the application. From source file:org.apache.flink.yarn.FlinkYarnClientBase.java
License:Apache License
/** * This method will block until the ApplicationMaster/JobManager have been * deployed on YARN./*w w w . j av a 2 s . c o m*/ */ protected AbstractFlinkYarnCluster deployInternal() throws Exception { isReadyForDeployment(); LOG.info("Using values:"); LOG.info("\tTaskManager count = {}", taskManagerCount); LOG.info("\tJobManager memory = {}", jobManagerMemoryMb); LOG.info("\tTaskManager memory = {}", taskManagerMemoryMb); // Create application via yarnClient yarnApplication = yarnClient.createApplication(); GetNewApplicationResponse appResponse = yarnApplication.getNewApplicationResponse(); // ------------------ Add dynamic properties to local flinkConfiguraton ------ Map<String, String> dynProperties = CliFrontend.getDynamicProperties(dynamicPropertiesEncoded); for (Map.Entry<String, String> dynProperty : dynProperties.entrySet()) { flinkConfiguration.setString(dynProperty.getKey(), dynProperty.getValue()); } 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); } // ------------------ Check if the specified queue exists -------------- try { List<QueueInfo> queues = yarnClient.getAllQueues(); if (queues.size() > 0 && this.yarnQueue != null) { // check only if there are queues configured in yarn and for this session. 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); } } // ------------------ Check if the YARN Cluster 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(); 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(); 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); } } // ------------------ Prepare Application Master Container ------------------------------ // respect custom JVM options in the YAML file final String javaOpts = flinkConfiguration.getString(ConfigConstants.FLINK_JVM_OPTIONS, ""); String logbackFile = configurationDirectory + File.separator + FlinkYarnSessionCli.CONFIG_FILE_LOGBACK_NAME; boolean hasLogback = new File(logbackFile).exists(); String log4jFile = configurationDirectory + File.separator + FlinkYarnSessionCli.CONFIG_FILE_LOG4J_NAME; boolean hasLog4j = new File(log4jFile).exists(); if (hasLogback) { shipFiles.add(new File(logbackFile)); } if (hasLog4j) { shipFiles.add(new File(log4jFile)); } // 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(jobManagerMemoryMb, flinkConfiguration) + "M " + javaOpts; if (hasLogback || hasLog4j) { amCommand += " -Dlog.file=\"" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/jobmanager.log\""; if (hasLogback) { amCommand += " -Dlogback.configurationFile=file:" + FlinkYarnSessionCli.CONFIG_FILE_LOGBACK_NAME; } if (hasLog4j) { amCommand += " -Dlog4j.configuration=file:" + FlinkYarnSessionCli.CONFIG_FILE_LOG4J_NAME; } } amCommand += " " + getApplicationMasterClass().getName() + " " + " 1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/jobmanager.out" + " 2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/jobmanager.err"; amContainer.setCommands(Collections.singletonList(amCommand)); LOG.debug("Application Master start command: " + amCommand); // intialize HDFS // 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"); } // Set-up ApplicationSubmissionContext for the application ApplicationSubmissionContext appContext = yarnApplication.getApplicationSubmissionContext(); 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)); } 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(fs, appId.toString(), flinkJarPath, appMasterJar, fs.getHomeDirectory()); Path remotePathConf = Utils.setupLocalResource(fs, appId.toString(), flinkConfigurationPath, flinkConf, fs.getHomeDirectory()); Map<String, LocalResource> localResources = new HashMap<>(2); localResources.put("flink.jar", appMasterJar); localResources.put("flink-conf.yaml", flinkConf); // setup security tokens (code from apache storm) final Path[] paths = new Path[2 + shipFiles.size()]; StringBuilder envShipFileList = new StringBuilder(); // 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[2 + i] = Utils.setupLocalResource(fs, appId.toString(), shipLocalPath, shipResources, fs.getHomeDirectory()); localResources.put(shipFile.getName(), shipResources); envShipFileList.append(paths[2 + i]); if (i + 1 < shipFiles.size()) { envShipFileList.append(','); } } paths[0] = remotePathJar; paths[1] = remotePathConf; 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. Utils.setTokensFor(amContainer, paths, conf); amContainer.setLocalResources(localResources); fs.close(); // Setup CLASSPATH for ApplicationMaster 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 classpath from YARN configuration Utils.setupEnv(conf, appMasterEnv); // 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)); if (dynamicPropertiesEncoded != null) { appMasterEnv.put(YarnConfigKeys.ENV_DYNAMIC_PROPERTIES, dynamicPropertiesEncoded); } 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 Runtime.getRuntime().addShutdownHook(deploymentFailureHook); LOG.info("Submitting application master " + appId); yarnClient.submitApplication(appContext); LOG.info("Waiting for the cluster to be allocated"); int waittime = 0; loop: while (true) { ApplicationReport report; 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: LOG.info("Deploying cluster, current state " + appState); if (waittime > 60000) { LOG.info( "Deployment took more than 60 seconds. Please check if the requested resources are available in the YARN cluster"); } } waittime += 1000; Thread.sleep(1000); } // print the application id for user to cancel themselves. if (isDetached()) { 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. } // the Flink cluster is deployed in YARN. Represent cluster return new FlinkYarnCluster(yarnClient, appId, conf, flinkConfiguration, sessionFilesDir, detached); }
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. * /*from w w w .j ava 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."); }//from w w w . j a va 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."); }/*from w w w . j av a 2 s. co m*/ ApplicationReport lastReport = null; if (pollingRunner == null) { LOG.warn("YarnClusterClient.getApplicationStatus() has been called on an uninitialized cluster." + "The system might be in an erroneous state"); } else { lastReport = pollingRunner.getLastReport(); } if (lastReport == null) { LOG.warn( "YarnClusterClient.getApplicationStatus() has been called on a cluster that didn't receive a status so far." + "The system might be in an erroneous state"); return ApplicationStatus.UNKNOWN; } else { YarnApplicationState appState = lastReport.getYarnApplicationState(); ApplicationStatus status = (appState == YarnApplicationState.FAILED || appState == YarnApplicationState.KILLED) ? ApplicationStatus.FAILED : ApplicationStatus.SUCCEEDED; if (status != ApplicationStatus.SUCCEEDED) { LOG.warn("YARN reported application state {}", appState); LOG.warn("Diagnostics: {}", lastReport.getDiagnostics()); } return status; } }
From source file:org.apache.flink.yarn.YarnClusterClient.java
License:Apache License
/** * Shuts down the Yarn application/*from www .j a v a 2s .c o m*/ */ public void shutdownCluster() { if (hasBeenShutDown.getAndSet(true)) { return; } if (!isConnected) { throw new IllegalStateException("The cluster has been not been connected to the ApplicationMaster."); } try { Runtime.getRuntime().removeShutdownHook(clientShutdownHook); } catch (IllegalStateException e) { // we are already in the shutdown hook } LOG.info("Sending shutdown request to the Application Master"); try { Future<Object> response = Patterns.ask(applicationClient.get(), new YarnMessages.LocalStopYarnSession( getApplicationStatus(), "Flink YARN Client requested shutdown"), new Timeout(akkaDuration)); Await.ready(response, akkaDuration); } catch (Exception e) { LOG.warn("Error while stopping YARN cluster.", e); } try { File propertiesFile = FlinkYarnSessionCli.getYarnPropertiesLocation(flinkConfig); if (propertiesFile.isFile()) { if (propertiesFile.delete()) { LOG.info("Deleted Yarn properties file at {}", propertiesFile.getAbsoluteFile().toString()); } else { LOG.warn("Couldn't delete Yarn properties file at {}", propertiesFile.getAbsoluteFile().toString()); } } } catch (Exception e) { LOG.warn("Exception while deleting the JobManager address file", e); } if (sessionFilesDir != null) { LOG.info("Deleting files in " + sessionFilesDir); try { FileSystem shutFS = FileSystem.get(hadoopConfig); shutFS.delete(sessionFilesDir, true); // delete conf and jar file. shutFS.close(); } catch (IOException e) { LOG.error("Could not delete the Flink jar and configuration files in HDFS..", e); } } else { LOG.warn("Session file directory not set. Not deleting session files"); } try { pollingRunner.stopRunner(); pollingRunner.join(1000); } catch (InterruptedException e) { LOG.warn("Shutdown of the polling runner was interrupted", e); Thread.currentThread().interrupt(); } try { ApplicationReport appReport = yarnClient.getApplicationReport(appId); LOG.info("Application " + appId + " finished with state " + appReport.getYarnApplicationState() + " and final state " + appReport.getFinalApplicationStatus() + " at " + appReport.getFinishTime()); if (appReport.getYarnApplicationState() == YarnApplicationState.FAILED || appReport.getYarnApplicationState() == YarnApplicationState.KILLED) { LOG.warn("Application failed. Diagnostics " + appReport.getDiagnostics()); LOG.warn("If log aggregation is activated in the Hadoop cluster, we recommend to retrieve " + "the full application log using this command:" + System.lineSeparator() + "\tyarn logs -applicationId " + appReport.getApplicationId() + System.lineSeparator() + "(It sometimes takes a few seconds until the logs are aggregated)"); } } catch (Exception e) { LOG.warn("Couldn't get final report", e); } LOG.info("YARN Client is shutting down"); yarnClient.stop(); // actorRunner is using the yarnClient. yarnClient = null; // set null to clearly see if somebody wants to access it afterwards. }
From source file:org.apache.flink.yarn.YarnClusterClientV2.java
License:Apache License
@Override protected JobSubmissionResult submitJob(JobGraph jobGraph, ClassLoader classLoader) throws ProgramInvocationException { try {//from ww w.j a va 2 s. c o m // Create application via yarnClient final YarnClientApplication yarnApplication = yarnClient.createApplication(); ApplicationReport report = this.clusterDescriptor.startAppMaster(jobGraph, yarnClient, yarnApplication); if (report.getYarnApplicationState().equals(YarnApplicationState.RUNNING)) { appId = report.getApplicationId(); trackingURL = report.getTrackingUrl(); logAndSysout("Please refer to " + getWebInterfaceURL() + " for the running status of job " + jobGraph.getJobID().toString()); //TODO: not support attach mode now return new JobSubmissionResult(jobGraph.getJobID()); } else { throw new ProgramInvocationException("Fail to submit the job."); } } catch (Exception e) { throw new ProgramInvocationException("Fail to submit the job", e.getCause()); } }
From source file:org.apache.flink.yarn.YARNSessionCapacitySchedulerITCase.java
License:Apache License
private void testDetachedPerJobYarnClusterInternal(String job) { YarnClient yc = YarnClient.createYarnClient(); yc.init(yarnConfiguration);/* w w w .j a v a 2 s. c om*/ yc.start(); // get temporary folder for writing output of wordcount example File tmpOutFolder = null; try { tmpOutFolder = tmp.newFolder(); } catch (IOException e) { throw new RuntimeException(e); } // get temporary file for reading input data for wordcount example File tmpInFile; try { tmpInFile = tmp.newFile(); FileUtils.writeStringToFile(tmpInFile, WordCountData.TEXT); } catch (IOException e) { throw new RuntimeException(e); } Runner runner = startWithArgs( new String[] { "run", "-m", "yarn-cluster", "-yj", flinkUberjar.getAbsolutePath(), "-yt", flinkLibFolder.getAbsolutePath(), "-yn", "1", "-yjm", "768", "-yD", "yarn.heap-cutoff-ratio=0.5", // test if the cutoff is passed correctly "-ytm", "1024", "-ys", "2", // test requesting slots from YARN. "--yarndetached", job, "--input", tmpInFile.getAbsoluteFile().toString(), "--output", tmpOutFolder.getAbsoluteFile().toString() }, "Job has been submitted with JobID", RunTypes.CLI_FRONTEND); // it should usually be 2, but on slow machines, the number varies Assert.assertTrue("There should be at most 2 containers running", getRunningContainers() <= 2); // give the runner some time to detach for (int attempt = 0; runner.isAlive() && attempt < 5; attempt++) { try { Thread.sleep(500); } catch (InterruptedException e) { } } Assert.assertFalse("The runner should detach.", runner.isAlive()); LOG.info("CLI Frontend has returned, so the job is running"); // find out the application id and wait until it has finished. try { List<ApplicationReport> apps = yc.getApplications(EnumSet.of(YarnApplicationState.RUNNING)); ApplicationId tmpAppId; if (apps.size() == 1) { // Better method to find the right appId. But sometimes the app is shutting down very fast // Only one running tmpAppId = apps.get(0).getApplicationId(); LOG.info("waiting for the job with appId {} to finish", tmpAppId); // wait until the app has finished while (yc.getApplications(EnumSet.of(YarnApplicationState.RUNNING)).size() > 0) { sleep(500); } } else { // get appId by finding the latest finished appid apps = yc.getApplications(); Collections.sort(apps, new Comparator<ApplicationReport>() { @Override public int compare(ApplicationReport o1, ApplicationReport o2) { return o1.getApplicationId().compareTo(o2.getApplicationId()) * -1; } }); tmpAppId = apps.get(0).getApplicationId(); LOG.info("Selected {} as the last appId from {}", tmpAppId, Arrays.toString(apps.toArray())); } final ApplicationId id = tmpAppId; // now it has finished. // check the output files. File[] listOfOutputFiles = tmpOutFolder.listFiles(); Assert.assertNotNull("Taskmanager output not found", listOfOutputFiles); LOG.info("The job has finished. TaskManager output files found in {}", tmpOutFolder); // read all output files in output folder to one output string String content = ""; for (File f : listOfOutputFiles) { if (f.isFile()) { content += FileUtils.readFileToString(f) + "\n"; } } //String content = FileUtils.readFileToString(taskmanagerOut); // check for some of the wordcount outputs. Assert.assertTrue("Expected string 'da 5' or '(all,2)' not found in string '" + content + "'", content.contains("da 5") || content.contains("(da,5)") || content.contains("(all,2)")); Assert.assertTrue("Expected string 'der 29' or '(mind,1)' not found in string'" + content + "'", content.contains("der 29") || content.contains("(der,29)") || content.contains("(mind,1)")); // check if the heap size for the TaskManager was set correctly File jobmanagerLog = YarnTestBase.findFile("..", new FilenameFilter() { @Override public boolean accept(File dir, String name) { return name.contains("jobmanager.log") && dir.getAbsolutePath().contains(id.toString()); } }); Assert.assertNotNull("Unable to locate JobManager log", jobmanagerLog); content = FileUtils.readFileToString(jobmanagerLog); // TM was started with 1024 but we cut off 50% (NOT THE DEFAULT VALUE) String expected = "Starting TaskManagers with command: $JAVA_HOME/bin/java -Xms424m -Xmx424m"; Assert.assertTrue( "Expected string '" + expected + "' not found in JobManager log: '" + jobmanagerLog + "'", content.contains(expected)); expected = " (2/2) (attempt #0) to "; Assert.assertTrue("Expected string '" + expected + "' not found in JobManager log." + "This string checks that the job has been started with a parallelism of 2. Log contents: '" + jobmanagerLog + "'", content.contains(expected)); // make sure the detached app is really finished. LOG.info("Checking again that app has finished"); ApplicationReport rep; do { sleep(500); rep = yc.getApplicationReport(id); LOG.info("Got report {}", rep); } while (rep.getYarnApplicationState() == YarnApplicationState.RUNNING); } catch (Throwable t) { LOG.warn("Error while detached yarn session was running", t); Assert.fail(t.getMessage()); } }
From source file:org.apache.flink.yarn.YarnTestBase.java
License:Apache License
@Before public void checkClusterEmpty() throws IOException, YarnException { if (yarnClient == null) { yarnClient = YarnClient.createYarnClient(); yarnClient.init(yarnConfiguration); yarnClient.start();/*from w w w . ja va 2 s. c o m*/ } List<ApplicationReport> apps = yarnClient.getApplications(); for (ApplicationReport app : apps) { if (app.getYarnApplicationState() != YarnApplicationState.FINISHED && app.getYarnApplicationState() != YarnApplicationState.KILLED && app.getYarnApplicationState() != YarnApplicationState.FAILED) { Assert.fail("There is at least one application on the cluster is not finished." + "App " + app.getApplicationId() + " is in state " + app.getYarnApplicationState()); } } }
From source file:org.apache.giraph.yarn.GiraphYarnClient.java
License:Apache License
/** * Assess whether job is already finished/failed and 'done' flag needs to be * set, prints progress display for client if all is going well. * @param report the application report to assess. * @return true if job report indicates the job run is over. */// w w w . j a v a 2 s. c om private boolean checkProgress(final ApplicationReport report) { YarnApplicationState jobState = report.getYarnApplicationState(); if (jobState == YarnApplicationState.FINISHED || jobState == YarnApplicationState.KILLED) { return true; } else if (jobState == YarnApplicationState.FAILED) { LOG.error(jobName + " reports FAILED state, diagnostics show: " + report.getDiagnostics()); return true; } else { if (reportCounter++ % 5 == 0) { displayJobReport(report); } } return false; }
From source file:org.apache.giraph.yarn.GiraphYarnClient.java
License:Apache License
/** * Display a formatted summary of the job progress report from the AM. * @param report the report to display.//from w ww . ja va 2 s. c om */ private void displayJobReport(final ApplicationReport report) { if (null == report) { throw new IllegalStateException( "[*] Latest ApplicationReport for job " + jobName + " was not received by the local client."); } final float elapsed = (System.currentTimeMillis() - report.getStartTime()) / 1000.0f; LOG.info(jobName + ", Elapsed: " + String.format("%.2f secs", elapsed)); LOG.info(report.getCurrentApplicationAttemptId() + ", State: " + report.getYarnApplicationState().name() + ", Containers used: " + report.getApplicationResourceUsageReport().getNumUsedContainers()); }