List of usage examples for org.apache.hadoop.yarn.api.records Container getNodeHttpAddress
@Public @Stable public abstract String getNodeHttpAddress();
From source file:org.apache.tez.dag.app.dag.impl.TestTaskAttempt.java
License:Apache License
@Test(timeout = 5000) // Ensure Container Preemption race with task completion is handled correctly by // the TaskAttempt public void testContainerPreemptedAfterSuccess() throws Exception { ApplicationId appId = ApplicationId.newInstance(1, 2); ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0); TezDAGID dagID = TezDAGID.getInstance(appId, 1); TezVertexID vertexID = TezVertexID.getInstance(dagID, 1); TezTaskID taskID = TezTaskID.getInstance(vertexID, 1); MockEventHandler eventHandler = spy(new MockEventHandler()); TaskAttemptListener taListener = mock(TaskAttemptListener.class); when(taListener.getAddress()).thenReturn(new InetSocketAddress("localhost", 0)); Configuration taskConf = new Configuration(); taskConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class); taskConf.setBoolean("fs.file.impl.disable.cache", true); TaskLocationHint locationHint = TaskLocationHint .createTaskLocationHint(new HashSet<String>(Arrays.asList(new String[] { "127.0.0.1" })), null); Resource resource = Resource.newInstance(1024, 1); NodeId nid = NodeId.newInstance("127.0.0.1", 0); ContainerId contId = ContainerId.newInstance(appAttemptId, 3); Container container = mock(Container.class); when(container.getId()).thenReturn(contId); when(container.getNodeId()).thenReturn(nid); when(container.getNodeHttpAddress()).thenReturn("localhost:0"); AppContext appCtx = mock(AppContext.class); AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskAttemptListener.class), new ContainerContextMatcher(), appCtx); containers.addContainerIfNew(container); doReturn(new ClusterInfo()).when(appCtx).getClusterInfo(); doReturn(containers).when(appCtx).getAllContainers(); TaskHeartbeatHandler mockHeartbeatHandler = mock(TaskHeartbeatHandler.class); TaskAttemptImpl taImpl = new MockTaskAttemptImpl(taskID, 1, eventHandler, taListener, taskConf, new SystemClock(), mockHeartbeatHandler, appCtx, locationHint, false, resource, createFakeContainerContext(), false); TezTaskAttemptID taskAttemptID = taImpl.getID(); ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class); taImpl.handle(new TaskAttemptEventSchedule(taskAttemptID, 0, 0)); // At state STARTING. taImpl.handle(new TaskAttemptEventStartedRemotely(taskAttemptID, contId, null)); assertEquals("Task attempt is not in the RUNNING state", taImpl.getState(), TaskAttemptState.RUNNING); verify(mockHeartbeatHandler).register(taskAttemptID); int expectedEventsAtRunning = 3; verify(eventHandler, times(expectedEventsAtRunning)).handle(arg.capture()); taImpl.handle(new TaskAttemptEvent(taskAttemptID, TaskAttemptEventType.TA_DONE)); assertEquals("Task attempt is not in the SUCCEEDED state", taImpl.getState(), TaskAttemptState.SUCCEEDED); verify(mockHeartbeatHandler).unregister(taskAttemptID); assertEquals(0, taImpl.getDiagnostics().size()); int expectedEventsAfterTerminating = expectedEventsAtRunning + 3; arg = ArgumentCaptor.forClass(Event.class); verify(eventHandler, times(expectedEventsAfterTerminating)).handle(arg.capture()); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEventsAfterTerminating), TaskEventTAUpdate.class, 1); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEventsAfterTerminating), AMSchedulerEventTAEnded.class, 1); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEventsAfterTerminating), DAGEventCounterUpdate.class, 1); taImpl.handle(new TaskAttemptEvent(taskAttemptID, TaskAttemptEventType.TA_CONTAINER_TERMINATED_BY_SYSTEM)); // verify unregister is not invoked again verify(mockHeartbeatHandler, times(1)).unregister(taskAttemptID); int expectedEventAfterTerminated = expectedEventsAfterTerminating + 0; arg = ArgumentCaptor.forClass(Event.class); verify(eventHandler, times(expectedEventAfterTerminated)).handle(arg.capture()); // Verify that the diagnostic message included in the Terminated event is not // captured - TA already succeeded. Error cause should be the default value assertEquals(0, taImpl.getDiagnostics().size()); assertEquals(TaskAttemptTerminationCause.UNKNOWN_ERROR, taImpl.getTerminationCause()); }
From source file:org.apache.tez.dag.app.dag.impl.TestTaskAttempt.java
License:Apache License
@Test(timeout = 5000) // Ensure node failure on Successful Non-Leaf tasks cause them to be marked as KILLED public void testNodeFailedNonLeafVertex() throws Exception { ApplicationId appId = ApplicationId.newInstance(1, 2); ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0); TezDAGID dagID = TezDAGID.getInstance(appId, 1); TezVertexID vertexID = TezVertexID.getInstance(dagID, 1); TezTaskID taskID = TezTaskID.getInstance(vertexID, 1); MockEventHandler eventHandler = spy(new MockEventHandler()); TaskAttemptListener taListener = mock(TaskAttemptListener.class); when(taListener.getAddress()).thenReturn(new InetSocketAddress("localhost", 0)); Configuration taskConf = new Configuration(); taskConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class); taskConf.setBoolean("fs.file.impl.disable.cache", true); TaskLocationHint locationHint = TaskLocationHint .createTaskLocationHint(new HashSet<String>(Arrays.asList(new String[] { "127.0.0.1" })), null); Resource resource = Resource.newInstance(1024, 1); NodeId nid = NodeId.newInstance("127.0.0.1", 0); ContainerId contId = ContainerId.newInstance(appAttemptId, 3); Container container = mock(Container.class); when(container.getId()).thenReturn(contId); when(container.getNodeId()).thenReturn(nid); when(container.getNodeHttpAddress()).thenReturn("localhost:0"); AppContext appCtx = mock(AppContext.class); AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskAttemptListener.class), new ContainerContextMatcher(), appCtx); containers.addContainerIfNew(container); doReturn(new ClusterInfo()).when(appCtx).getClusterInfo(); doReturn(containers).when(appCtx).getAllContainers(); TaskHeartbeatHandler mockHeartbeatHandler = mock(TaskHeartbeatHandler.class); MockTaskAttemptImpl taImpl = new MockTaskAttemptImpl(taskID, 1, eventHandler, taListener, taskConf, new SystemClock(), mockHeartbeatHandler, appCtx, locationHint, false, resource, createFakeContainerContext(), false); TezTaskAttemptID taskAttemptID = taImpl.getID(); ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class); taImpl.handle(new TaskAttemptEventSchedule(taskAttemptID, 0, 0)); // At state STARTING. taImpl.handle(new TaskAttemptEventStartedRemotely(taskAttemptID, contId, null)); assertEquals("Task attempt is not in the RUNNING state", TaskAttemptState.RUNNING, taImpl.getState()); verify(mockHeartbeatHandler).register(taskAttemptID); int expectedEventsAtRunning = 3; verify(eventHandler, times(expectedEventsAtRunning)).handle(arg.capture()); taImpl.handle(new TaskAttemptEvent(taskAttemptID, TaskAttemptEventType.TA_DONE)); assertEquals("Task attempt is not in the SUCCEEDED state", TaskAttemptState.SUCCEEDED, taImpl.getState()); verify(mockHeartbeatHandler).unregister(taskAttemptID); assertEquals(0, taImpl.getDiagnostics().size()); int expectedEvenstAfterTerminating = expectedEventsAtRunning + 3; arg = ArgumentCaptor.forClass(Event.class); verify(eventHandler, times(expectedEvenstAfterTerminating)).handle(arg.capture()); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), TaskEventTAUpdate.class, 1); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), AMSchedulerEventTAEnded.class, 1); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), DAGEventCounterUpdate.class, 1); // Send out a Node Failure. taImpl.handle(new TaskAttemptEventNodeFailed(taskAttemptID, "NodeDecomissioned", TaskAttemptTerminationCause.NODE_FAILED)); // Verify in KILLED state assertEquals("Task attempt is not in the KILLED state", TaskAttemptState.KILLED, taImpl.getState()); // verify unregister is not invoked again verify(mockHeartbeatHandler, times(1)).unregister(taskAttemptID); assertEquals(true, taImpl.inputFailedReported); // Verify one event to the Task informing it about FAILURE. No events to scheduler. Counter event. int expectedEventsNodeFailure = expectedEvenstAfterTerminating + 2; arg = ArgumentCaptor.forClass(Event.class); verify(eventHandler, times(expectedEventsNodeFailure)).handle(arg.capture()); verifyEventType(arg.getAllValues().subList(expectedEvenstAfterTerminating, expectedEventsNodeFailure), TaskEventTAUpdate.class, 1); // Verify still in KILLED state assertEquals("Task attempt is not in the KILLED state", TaskAttemptState.KILLED, taImpl.getState()); assertEquals(TaskAttemptTerminationCause.NODE_FAILED, taImpl.getTerminationCause()); }
From source file:org.apache.tez.dag.app.dag.impl.TestTaskAttempt.java
License:Apache License
@Test(timeout = 5000) // Ensure node failure on Successful Leaf tasks do not cause them to be marked as KILLED public void testNodeFailedLeafVertex() throws Exception { ApplicationId appId = ApplicationId.newInstance(1, 2); ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0); TezDAGID dagID = TezDAGID.getInstance(appId, 1); TezVertexID vertexID = TezVertexID.getInstance(dagID, 1); TezTaskID taskID = TezTaskID.getInstance(vertexID, 1); MockEventHandler eventHandler = spy(new MockEventHandler()); TaskAttemptListener taListener = mock(TaskAttemptListener.class); when(taListener.getAddress()).thenReturn(new InetSocketAddress("localhost", 0)); Configuration taskConf = new Configuration(); taskConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class); taskConf.setBoolean("fs.file.impl.disable.cache", true); TaskLocationHint locationHint = TaskLocationHint .createTaskLocationHint(new HashSet<String>(Arrays.asList(new String[] { "127.0.0.1" })), null); Resource resource = Resource.newInstance(1024, 1); NodeId nid = NodeId.newInstance("127.0.0.1", 0); ContainerId contId = ContainerId.newInstance(appAttemptId, 3); Container container = mock(Container.class); when(container.getId()).thenReturn(contId); when(container.getNodeId()).thenReturn(nid); when(container.getNodeHttpAddress()).thenReturn("localhost:0"); AppContext appCtx = mock(AppContext.class); AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskAttemptListener.class), new ContainerContextMatcher(), appCtx); containers.addContainerIfNew(container); doReturn(new ClusterInfo()).when(appCtx).getClusterInfo(); doReturn(containers).when(appCtx).getAllContainers(); TaskHeartbeatHandler mockHeartbeatHandler = mock(TaskHeartbeatHandler.class); TaskAttemptImpl taImpl = new MockTaskAttemptImpl(taskID, 1, eventHandler, taListener, taskConf, new SystemClock(), mockHeartbeatHandler, appCtx, locationHint, false, resource, createFakeContainerContext(), true); TezTaskAttemptID taskAttemptID = taImpl.getID(); ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class); taImpl.handle(new TaskAttemptEventSchedule(taskAttemptID, 0, 0)); // At state STARTING. taImpl.handle(new TaskAttemptEventStartedRemotely(taskAttemptID, contId, null)); assertEquals("Task attempt is not in the RUNNING state", TaskAttemptState.RUNNING, taImpl.getState()); verify(mockHeartbeatHandler).register(taskAttemptID); int expectedEventsAtRunning = 3; verify(eventHandler, times(expectedEventsAtRunning)).handle(arg.capture()); taImpl.handle(new TaskAttemptEvent(taskAttemptID, TaskAttemptEventType.TA_DONE)); assertEquals("Task attempt is not in the SUCCEEDED state", TaskAttemptState.SUCCEEDED, taImpl.getState()); verify(mockHeartbeatHandler).unregister(taskAttemptID); assertEquals(0, taImpl.getDiagnostics().size()); int expectedEvenstAfterTerminating = expectedEventsAtRunning + 3; arg = ArgumentCaptor.forClass(Event.class); verify(eventHandler, times(expectedEvenstAfterTerminating)).handle(arg.capture()); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), TaskEventTAUpdate.class, 1); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), AMSchedulerEventTAEnded.class, 1); verifyEventType(arg.getAllValues().subList(expectedEventsAtRunning, expectedEvenstAfterTerminating), DAGEventCounterUpdate.class, 1); // Send out a Node Failure. taImpl.handle(new TaskAttemptEventNodeFailed(taskAttemptID, "NodeDecomissioned", TaskAttemptTerminationCause.NODE_FAILED)); // Verify no additional events int expectedEventsNodeFailure = expectedEvenstAfterTerminating + 0; arg = ArgumentCaptor.forClass(Event.class); verify(eventHandler, times(expectedEventsNodeFailure)).handle(arg.capture()); // Verify still in SUCCEEDED state assertEquals("Task attempt is not in the SUCCEEDED state", TaskAttemptState.SUCCEEDED, taImpl.getState()); // verify unregister is not invoked again verify(mockHeartbeatHandler, times(1)).unregister(taskAttemptID); // error cause remains as default value assertEquals(TaskAttemptTerminationCause.UNKNOWN_ERROR, taImpl.getTerminationCause()); }
From source file:org.apache.tez.dag.app.dag.impl.TestTaskAttempt.java
License:Apache License
@Test(timeout = 5000) // Verifies that multiple TooManyFetchFailures are handled correctly by the // TaskAttempt.//from w ww .j av a 2 s. c o m public void testMultipleOutputFailed() throws Exception { ApplicationId appId = ApplicationId.newInstance(1, 2); ApplicationAttemptId appAttemptId = ApplicationAttemptId.newInstance(appId, 0); TezDAGID dagID = TezDAGID.getInstance(appId, 1); TezVertexID vertexID = TezVertexID.getInstance(dagID, 1); TezTaskID taskID = TezTaskID.getInstance(vertexID, 1); MockEventHandler mockEh = new MockEventHandler(); MockEventHandler eventHandler = spy(mockEh); TaskAttemptListener taListener = mock(TaskAttemptListener.class); when(taListener.getAddress()).thenReturn(new InetSocketAddress("localhost", 0)); Configuration taskConf = new Configuration(); taskConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class); taskConf.setBoolean("fs.file.impl.disable.cache", true); TaskLocationHint locationHint = TaskLocationHint .createTaskLocationHint(new HashSet<String>(Arrays.asList(new String[] { "127.0.0.1" })), null); Resource resource = Resource.newInstance(1024, 1); NodeId nid = NodeId.newInstance("127.0.0.1", 0); ContainerId contId = ContainerId.newInstance(appAttemptId, 3); Container container = mock(Container.class); when(container.getId()).thenReturn(contId); when(container.getNodeId()).thenReturn(nid); when(container.getNodeHttpAddress()).thenReturn("localhost:0"); AppContext appCtx = mock(AppContext.class); AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskAttemptListener.class), new ContainerContextMatcher(), appCtx); containers.addContainerIfNew(container); doReturn(new ClusterInfo()).when(appCtx).getClusterInfo(); doReturn(containers).when(appCtx).getAllContainers(); TaskHeartbeatHandler mockHeartbeatHandler = mock(TaskHeartbeatHandler.class); MockTaskAttemptImpl taImpl = new MockTaskAttemptImpl(taskID, 1, eventHandler, taListener, taskConf, new SystemClock(), mockHeartbeatHandler, appCtx, locationHint, false, resource, createFakeContainerContext(), false); TezTaskAttemptID taskAttemptID = taImpl.getID(); taImpl.handle(new TaskAttemptEventSchedule(taskAttemptID, 0, 0)); // At state STARTING. taImpl.handle(new TaskAttemptEventStartedRemotely(taskAttemptID, contId, null)); verify(mockHeartbeatHandler).register(taskAttemptID); taImpl.handle(new TaskAttemptEvent(taskAttemptID, TaskAttemptEventType.TA_DONE)); assertEquals("Task attempt is not in succeeded state", taImpl.getState(), TaskAttemptState.SUCCEEDED); verify(mockHeartbeatHandler).unregister(taskAttemptID); int expectedEventsTillSucceeded = 6; ArgumentCaptor<Event> arg = ArgumentCaptor.forClass(Event.class); verify(eventHandler, times(expectedEventsTillSucceeded)).handle(arg.capture()); verifyEventType(arg.getAllValues(), TaskEventTAUpdate.class, 2); InputReadErrorEvent mockReEvent = InputReadErrorEvent.create("", 0, 1); EventMetaData mockMeta = mock(EventMetaData.class); TezTaskAttemptID mockDestId1 = mock(TezTaskAttemptID.class); when(mockMeta.getTaskAttemptID()).thenReturn(mockDestId1); TezEvent tzEvent = new TezEvent(mockReEvent, mockMeta); taImpl.handle(new TaskAttemptEventOutputFailed(taskAttemptID, tzEvent, 4)); // failure threshold not met. state is SUCCEEDED assertEquals("Task attempt is not in succeeded state", taImpl.getState(), TaskAttemptState.SUCCEEDED); // sending same error again doesnt change anything taImpl.handle(new TaskAttemptEventOutputFailed(taskAttemptID, tzEvent, 4)); assertEquals("Task attempt is not in succeeded state", taImpl.getState(), TaskAttemptState.SUCCEEDED); // default value of error cause assertEquals(TaskAttemptTerminationCause.UNKNOWN_ERROR, taImpl.getTerminationCause()); // different destination attempt reports error. now threshold crossed TezTaskAttemptID mockDestId2 = mock(TezTaskAttemptID.class); when(mockMeta.getTaskAttemptID()).thenReturn(mockDestId2); taImpl.handle(new TaskAttemptEventOutputFailed(taskAttemptID, tzEvent, 4)); assertEquals("Task attempt is not in FAILED state", taImpl.getState(), TaskAttemptState.FAILED); assertEquals(TaskAttemptTerminationCause.OUTPUT_LOST, taImpl.getTerminationCause()); // verify unregister is not invoked again verify(mockHeartbeatHandler, times(1)).unregister(taskAttemptID); assertEquals(true, taImpl.inputFailedReported); int expectedEventsAfterFetchFailure = expectedEventsTillSucceeded + 2; arg.getAllValues().clear(); verify(eventHandler, times(expectedEventsAfterFetchFailure)).handle(arg.capture()); verifyEventType(arg.getAllValues().subList(expectedEventsTillSucceeded, expectedEventsAfterFetchFailure), TaskEventTAUpdate.class, 1); taImpl.handle(new TaskAttemptEventOutputFailed(taskAttemptID, tzEvent, 1)); assertEquals("Task attempt is not in FAILED state, still", taImpl.getState(), TaskAttemptState.FAILED); assertFalse("InternalError occurred trying to handle TA_TOO_MANY_FETCH_FAILURES", eventHandler.internalError); // No new events. verify(eventHandler, times(expectedEventsAfterFetchFailure)).handle(arg.capture()); }
From source file:org.apache.tez.dag.app.dag.impl.TestVertexImpl.java
License:Apache License
@SuppressWarnings("unchecked") @Test(timeout = 5000)//from w w w. ja v a 2 s. com public void testVertexTaskAttemptProcessorFailure() throws Exception { initAllVertices(VertexState.INITED); VertexImpl v = vertices.get("vertex1"); startVertex(v); dispatcher.await(); TaskAttemptImpl ta = (TaskAttemptImpl) v.getTask(0).getAttempts().values().iterator().next(); ta.handle(new TaskAttemptEventSchedule(ta.getID(), 2, 2)); NodeId nid = NodeId.newInstance("127.0.0.1", 0); ContainerId contId = ContainerId.newInstance(appAttemptId, 3); Container container = mock(Container.class); when(container.getId()).thenReturn(contId); when(container.getNodeId()).thenReturn(nid); when(container.getNodeHttpAddress()).thenReturn("localhost:0"); AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskAttemptListener.class), new ContainerContextMatcher(), appContext); containers.addContainerIfNew(container); doReturn(containers).when(appContext).getAllContainers(); ta.handle(new TaskAttemptEventStartedRemotely(ta.getID(), contId, null)); Assert.assertEquals(TaskAttemptStateInternal.RUNNING, ta.getInternalState()); dispatcher.getEventHandler() .handle(new VertexEventRouteEvent(v.getVertexId(), Collections.singletonList(new TezEvent( new TaskAttemptFailedEvent("Failed"), new EventMetaData(EventProducerConsumerType.PROCESSOR, v.getName(), null, ta.getID()))))); dispatcher.await(); Assert.assertEquals(VertexState.RUNNING, v.getState()); Assert.assertEquals(TaskAttemptTerminationCause.APPLICATION_ERROR, ta.getTerminationCause()); }
From source file:org.apache.tez.dag.app.dag.impl.TestVertexImpl.java
License:Apache License
@SuppressWarnings("unchecked") @Test(timeout = 5000)//from w w w.j a va2 s.c om public void testVertexTaskAttemptInputFailure() throws Exception { initAllVertices(VertexState.INITED); VertexImpl v = vertices.get("vertex1"); startVertex(v); dispatcher.await(); TaskAttemptImpl ta = (TaskAttemptImpl) v.getTask(0).getAttempts().values().iterator().next(); ta.handle(new TaskAttemptEventSchedule(ta.getID(), 2, 2)); NodeId nid = NodeId.newInstance("127.0.0.1", 0); ContainerId contId = ContainerId.newInstance(appAttemptId, 3); Container container = mock(Container.class); when(container.getId()).thenReturn(contId); when(container.getNodeId()).thenReturn(nid); when(container.getNodeHttpAddress()).thenReturn("localhost:0"); AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskAttemptListener.class), new ContainerContextMatcher(), appContext); containers.addContainerIfNew(container); doReturn(containers).when(appContext).getAllContainers(); ta.handle(new TaskAttemptEventStartedRemotely(ta.getID(), contId, null)); Assert.assertEquals(TaskAttemptStateInternal.RUNNING, ta.getInternalState()); dispatcher.getEventHandler() .handle(new VertexEventRouteEvent(v.getVertexId(), Collections.singletonList(new TezEvent( new TaskAttemptFailedEvent("Failed"), new EventMetaData(EventProducerConsumerType.INPUT, v.getName(), null, ta.getID()))))); dispatcher.await(); Assert.assertEquals(VertexState.RUNNING, v.getState()); Assert.assertEquals(TaskAttemptTerminationCause.INPUT_READ_ERROR, ta.getTerminationCause()); }
From source file:org.apache.tez.dag.app.dag.impl.TestVertexImpl.java
License:Apache License
@SuppressWarnings("unchecked") @Test(timeout = 5000)//w w w . ja va 2 s .c o m public void testVertexTaskAttemptOutputFailure() throws Exception { initAllVertices(VertexState.INITED); VertexImpl v = vertices.get("vertex1"); startVertex(v); dispatcher.await(); TaskAttemptImpl ta = (TaskAttemptImpl) v.getTask(0).getAttempts().values().iterator().next(); ta.handle(new TaskAttemptEventSchedule(ta.getID(), 2, 2)); NodeId nid = NodeId.newInstance("127.0.0.1", 0); ContainerId contId = ContainerId.newInstance(appAttemptId, 3); Container container = mock(Container.class); when(container.getId()).thenReturn(contId); when(container.getNodeId()).thenReturn(nid); when(container.getNodeHttpAddress()).thenReturn("localhost:0"); AMContainerMap containers = new AMContainerMap(mock(ContainerHeartbeatHandler.class), mock(TaskAttemptListener.class), new ContainerContextMatcher(), appContext); containers.addContainerIfNew(container); doReturn(containers).when(appContext).getAllContainers(); ta.handle(new TaskAttemptEventStartedRemotely(ta.getID(), contId, null)); Assert.assertEquals(TaskAttemptStateInternal.RUNNING, ta.getInternalState()); dispatcher.getEventHandler() .handle(new VertexEventRouteEvent(v.getVertexId(), Collections.singletonList(new TezEvent( new TaskAttemptFailedEvent("Failed"), new EventMetaData(EventProducerConsumerType.OUTPUT, v.getName(), null, ta.getID()))))); dispatcher.await(); Assert.assertEquals(VertexState.RUNNING, v.getState()); Assert.assertEquals(TaskAttemptTerminationCause.OUTPUT_WRITE_ERROR, ta.getTerminationCause()); }
From source file:org.hdl.caffe.yarn.app.ApplicationMaster.java
License:Apache License
public boolean startAllContainers() throws Exception { if (numAllocatedContainers.get() == numTotalContainers) { int numCaffeContainers = 0; if (this.allocatedContainers.size() < numTotalProcessorContainers) { LOG.error("not enough caffe containers allocated!"); return false; }/*w ww . ja va2 s . c o m*/ for (Container allocatedContainer : this.allocatedContainers) { if (numCaffeContainers < numTotalProcessorContainers) { LOG.info("work cid: " + allocatedContainer.getId().toString()); clusterSpec.addServerSpec(allocatedContainer.getId().toString(), allocatedContainer.getNodeId().getHost()); numCaffeContainers++; } else { break; } } for (Container allocatedContainer : this.allocatedContainers) { LOG.info("Launching a new container." + ", containerId=" + allocatedContainer.getId() + ", containerNode=" + allocatedContainer.getNodeId().getHost() + ":" + allocatedContainer.getNodeId().getPort() + ", containerNodeURI=" + allocatedContainer.getNodeHttpAddress() + ", containerResourceMemory" + allocatedContainer.getResource().getMemorySize() + ", containerResourceVirtualCores" + allocatedContainer.getResource().getVirtualCores()); LOG.info("server cid: " + allocatedContainer.getId().toString()); LaunchContainerThread launchDelegator = new LaunchContainerThread(allocatedContainer, train, solver, feature, label, model, output, connection, this, clusterSpec.getServerAddress(allocatedContainer.getId().toString())); LOG.info("ServerAdress: " + clusterSpec.getServerAddress(allocatedContainer.getId().toString())); launchDelegator.setCaffeProcessorJar(caffeProcessorJar); launchDelegator.setContainerMemory(containerMemory); launchDelegator.setContainerRetryPolicy(containerRetryPolicy); launchDelegator.setContainerRetryErrorCodes(containerRetryErrorCodes); launchDelegator.setContainerMaxRetries(containerMaxRetries); launchDelegator.setContainrRetryInterval(containrRetryInterval); Thread launchThread = new Thread(launchDelegator); // launch and start the container on a separate thread to keep // the main thread unblocked // as all containers may not be allocated at one go. launchThreads.add(launchThread); launchedContainers.add(allocatedContainer.getId()); launchThread.start(); } } else { throw new Exception("containers are not allocated!"); } return true; }
From source file:tachyon.yarn.ApplicationMaster.java
License:Apache License
private void launchTachyonMasterContainers(List<Container> containers) { final String command = new CommandBuilder( PathUtils.concatPath(mTachyonHome, "integration", "bin", "tachyon-master-yarn.sh")) .addArg("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stdout") .addArg("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stderr").toString(); List<String> commands = Lists.newArrayList(command); for (Container container : containers) { try {/*from w ww .j a v a 2s . c o m*/ // Launch container by create ContainerLaunchContext ContainerLaunchContext ctx = Records.newRecord(ContainerLaunchContext.class); ctx.setCommands(commands); LOG.info("Launching container {} for Tachyon master on {} with master command: {}", container.getId(), container.getNodeHttpAddress(), commands); mNMClient.startContainer(container, ctx); String containerUri = container.getNodeHttpAddress(); // in the form of 1.2.3.4:8042 mMasterContainerNetAddress = containerUri.split(":")[0]; LOG.info("Master address: " + mMasterContainerNetAddress); mMasterContainerAllocated = true; return; } catch (Exception ex) { LOG.error("Error launching container " + container.getId() + " " + ex); } } }
From source file:tachyon.yarn.ApplicationMaster.java
License:Apache License
private void launchTachyonWorkerContainers(List<Container> containers) { final String command = new CommandBuilder( PathUtils.concatPath(mTachyonHome, "integration", "bin", "tachyon-worker-yarn.sh")) .addArg("1>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stdout") .addArg("2>" + ApplicationConstants.LOG_DIR_EXPANSION_VAR + "/stderr").toString(); List<String> commands = Lists.newArrayList(command); Map<String, String> environmentMap = new HashMap<String, String>(); environmentMap.put("TACHYON_MASTER_ADDRESS", mMasterContainerNetAddress); environmentMap.put("TACHYON_WORKER_MEMORY_SIZE", FormatUtils.getSizeFromBytes((long) mRamdiskMemInMB * Constants.MB)); for (Container container : containers) { if (mNumAllocatedWorkerContainers >= mNumWorkers) { break; }// w w w . jav a2 s. c o m try { // Launch container by create ContainerLaunchContext ContainerLaunchContext ctx = Records.newRecord(ContainerLaunchContext.class); ctx.setCommands(commands); ctx.setEnvironment(environmentMap); LOG.info("Launching container {} for Tachyon worker {} on {} with worker command: {}", container.getId(), mNumAllocatedWorkerContainers, container.getNodeHttpAddress(), command); mNMClient.startContainer(container, ctx); mNumAllocatedWorkerContainers++; } catch (Exception ex) { LOG.error("Error launching container " + container.getId() + " " + ex); } } }