Apache
/
hadoop
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

MAPREDUCE-5409. MRAppMaster throws InvalidStateTransitonException: Invalid event: TA_TOO_MANY_FETCH_FAILURE at KILLED for TaskAttemptImpl. Contributed by Gera Shegalov 

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1548197 13f79535-47bb-0310-9956-ffa450edef68
This commit is contained in:
Jason Darrell Lowe 2013-12-05 16:30:12 +00:00
parent 13c5d3eb34
commit 0f2ce45aeb
3 changed files with 89 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
hadoop-mapreduce-project/CHANGES.txt
View File

@ -280,6 +280,10 @@ Release 2.3.0 - UNRELEASED
MAPREDUCE-5451. MR uses LD_LIBRARY_PATH which doesn't mean anything in
Windows. (Yingda Chen via cnauroth)
MAPREDUCE-5409. MRAppMaster throws InvalidStateTransitonException: Invalid
event: TA_TOO_MANY_FETCH_FAILURE at KILLED for TaskAttemptImpl (Gera
Shegalov via jlowe)
Release 2.2.0 - 2013-10-13
INCOMPATIBLE CHANGES

40
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TaskAttemptImpl.java
View File

@ -192,6 +192,21 @@ public abstract class TaskAttemptImpl implements
DIAGNOSTIC_INFORMATION_UPDATE_TRANSITION
= new DiagnosticInformationUpdater();
private static final EnumSet<TaskAttemptEventType>
FAILED_KILLED_STATE_IGNORED_EVENTS = EnumSet.of(
TaskAttemptEventType.TA_KILL,
TaskAttemptEventType.TA_ASSIGNED,
TaskAttemptEventType.TA_CONTAINER_COMPLETED,
TaskAttemptEventType.TA_UPDATE,
// Container launch events can arrive late
TaskAttemptEventType.TA_CONTAINER_LAUNCHED,
TaskAttemptEventType.TA_CONTAINER_LAUNCH_FAILED,
TaskAttemptEventType.TA_CONTAINER_CLEANED,
TaskAttemptEventType.TA_COMMIT_PENDING,
TaskAttemptEventType.TA_DONE,
TaskAttemptEventType.TA_FAILMSG,
TaskAttemptEventType.TA_TOO_MANY_FETCH_FAILURE);
private static final StateMachineFactory
<TaskAttemptImpl, TaskAttemptStateInternal, TaskAttemptEventType, TaskAttemptEvent>
stateMachineFactory
@ -452,18 +467,7 @@ public abstract class TaskAttemptImpl implements
DIAGNOSTIC_INFORMATION_UPDATE_TRANSITION)
// Ignore-able events for FAILED state
.addTransition(TaskAttemptStateInternal.FAILED, TaskAttemptStateInternal.FAILED,
EnumSet.of(TaskAttemptEventType.TA_KILL,
TaskAttemptEventType.TA_ASSIGNED,
TaskAttemptEventType.TA_CONTAINER_COMPLETED,
TaskAttemptEventType.TA_UPDATE,
// Container launch events can arrive late
TaskAttemptEventType.TA_CONTAINER_LAUNCHED,
TaskAttemptEventType.TA_CONTAINER_LAUNCH_FAILED,
TaskAttemptEventType.TA_CONTAINER_CLEANED,
TaskAttemptEventType.TA_COMMIT_PENDING,
TaskAttemptEventType.TA_DONE,
TaskAttemptEventType.TA_FAILMSG,
TaskAttemptEventType.TA_TOO_MANY_FETCH_FAILURE))
FAILED_KILLED_STATE_IGNORED_EVENTS)
// Transitions from KILLED state
.addTransition(TaskAttemptStateInternal.KILLED, TaskAttemptStateInternal.KILLED,
@ -471,17 +475,7 @@ public abstract class TaskAttemptImpl implements
DIAGNOSTIC_INFORMATION_UPDATE_TRANSITION)
// Ignore-able events for KILLED state
.addTransition(TaskAttemptStateInternal.KILLED, TaskAttemptStateInternal.KILLED,
EnumSet.of(TaskAttemptEventType.TA_KILL,
TaskAttemptEventType.TA_ASSIGNED,
TaskAttemptEventType.TA_CONTAINER_COMPLETED,
TaskAttemptEventType.TA_UPDATE,
// Container launch events can arrive late
TaskAttemptEventType.TA_CONTAINER_LAUNCHED,
TaskAttemptEventType.TA_CONTAINER_LAUNCH_FAILED,
TaskAttemptEventType.TA_CONTAINER_CLEANED,
TaskAttemptEventType.TA_COMMIT_PENDING,
TaskAttemptEventType.TA_DONE,
TaskAttemptEventType.TA_FAILMSG))
FAILED_KILLED_STATE_IGNORED_EVENTS)
// create the topology tables
.installTopology();

68
hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestTaskAttempt.java
View File

@ -550,6 +550,8 @@ public class TestTaskAttempt{
eventHandler.internalError);
}
@Test
public void testAppDiognosticEventOnUnassignedTask() throws Exception {
ApplicationId appId = ApplicationId.newInstance(1, 2);
@ -599,6 +601,72 @@ public class TestTaskAttempt{
eventHandler.internalError);
}
@Test
public void testTooManyFetchFailureAfterKill() throws Exception {
ApplicationId appId = ApplicationId.newInstance(1, 2);
ApplicationAttemptId appAttemptId =
ApplicationAttemptId.newInstance(appId, 0);
JobId jobId = MRBuilderUtils.newJobId(appId, 1);
TaskId taskId = MRBuilderUtils.newTaskId(jobId, 1, TaskType.MAP);
TaskAttemptId attemptId = MRBuilderUtils.newTaskAttemptId(taskId, 0);
Path jobFile = mock(Path.class);
MockEventHandler eventHandler = new MockEventHandler();
TaskAttemptListener taListener = mock(TaskAttemptListener.class);
when(taListener.getAddress()).thenReturn(new InetSocketAddress("localhost", 0));
JobConf jobConf = new JobConf();
jobConf.setClass("fs.file.impl", StubbedFS.class, FileSystem.class);
jobConf.setBoolean("fs.file.impl.disable.cache", true);
jobConf.set(JobConf.MAPRED_MAP_TASK_ENV, "");
jobConf.set(MRJobConfig.APPLICATION_ATTEMPT_ID, "10");
TaskSplitMetaInfo splits = mock(TaskSplitMetaInfo.class);
when(splits.getLocations()).thenReturn(new String[] {"127.0.0.1"});
AppContext appCtx = mock(AppContext.class);
ClusterInfo clusterInfo = mock(ClusterInfo.class);
Resource resource = mock(Resource.class);
when(appCtx.getClusterInfo()).thenReturn(clusterInfo);
when(resource.getMemory()).thenReturn(1024);
TaskAttemptImpl taImpl =
new MapTaskAttemptImpl(taskId, 1, eventHandler, jobFile, 1,
splits, jobConf, taListener,
mock(Token.class), new Credentials(),
new SystemClock(), appCtx);
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");
taImpl.handle(new TaskAttemptEvent(attemptId,
TaskAttemptEventType.TA_SCHEDULE));
taImpl.handle(new TaskAttemptContainerAssignedEvent(attemptId,
container, mock(Map.class)));
taImpl.handle(new TaskAttemptContainerLaunchedEvent(attemptId, 0));
taImpl.handle(new TaskAttemptEvent(attemptId,
TaskAttemptEventType.TA_DONE));
taImpl.handle(new TaskAttemptEvent(attemptId,
TaskAttemptEventType.TA_CONTAINER_CLEANED));
assertEquals("Task attempt is not in succeeded state", taImpl.getState(),
TaskAttemptState.SUCCEEDED);
taImpl.handle(new TaskAttemptEvent(attemptId,
TaskAttemptEventType.TA_KILL));
assertEquals("Task attempt is not in KILLED state", taImpl.getState(),
TaskAttemptState.KILLED);
taImpl.handle(new TaskAttemptEvent(attemptId,
TaskAttemptEventType.TA_TOO_MANY_FETCH_FAILURE));
assertEquals("Task attempt is not in KILLED state, still", taImpl.getState(),
TaskAttemptState.KILLED);
assertFalse("InternalError occurred trying to handle TA_CONTAINER_CLEANED",
eventHandler.internalError);
}
@Test
public void testAppDiognosticEventOnNewTask() throws Exception {
ApplicationId appId = ApplicationId.newInstance(1, 2);