We are running same job all the time. And that error is happening from time to time.
Here is job submittion code: private JobSubmissionResult submitProgramToCluster(PackagedProgram packagedProgram) throws JobSubmitterException, ProgramMissingJobException, ProgramInvocationException { ClusterClient clusterClient = clusterClientUtil.getPrototypeClusterClient(); int parallelism = Integer.parseInt(serverConfiguration.envParallelism); return clusterClient.run(packagedProgram, parallelism); } And here our util for retrieving ClusterClient. public class ClusterClientUtil { ... public ClusterClient getPrototypeClusterClient() throws JobSubmitterException { return createClusterClientInstance(); } private synchronized ClusterClient createClusterClientInstance() throws JobSubmitterException { try { LOG.info("Creating new ClusterClient instance."); Configuration configuration = flinkConfigurator.getFlinkConfiguration(); ApplicationId applicationId = ConverterUtils.toApplicationId(configuration.getString(FlinkConfigurator.PROPERTY_FLINK_APP_ID, "")); logger.debug("Retrieved Flink applicationId: {}", applicationId.toString()); YarnClusterDescriptor yarnClusterDescriptor = new YarnClusterDescriptor(configuration, serverConfiguration.getFlinkConfigurationDirectory); ApplicationReport applicationReport = yarnClient.getApplicationReport(applicationId); final int numberTaskManagers = configuration.getInteger(FlinkConfigurator.PROPERTY_FLINK_NUMBER_TASK_MANAGERS, 0); final int slotsPerTaskManager = configuration.getInteger(FlinkConfigurator.PROPERTY_FLINK_SLOTS_PER_TASK_MANAGER, -1); return new YarnClusterClient(yarnClusterDescriptor, numberTaskManagers, slotsPerTaskManager, yarnClient, applicationReport, configuration, false); } catch (Exception e) { throw new JobSubmitterException("Unable to create YarnClusterClient.", e); } } What Yarn settings to you need? -- Sent from: http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/