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

MAPREDUCE-7353: Mapreduce job fails when NM is stopped. Contributed by Bilwa S T (BilwaST) 

(cherry picked from commit 7581413156)
This commit is contained in:
Eric Payne 2021-07-07 20:43:44 +00:00
parent 99cff01e38
commit e395711164
2 changed files with 36 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
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

@ -492,10 +492,15 @@ public abstract class TaskAttemptImpl implements
TaskAttemptStateInternal.SUCCESS_CONTAINER_CLEANUP,
TaskAttemptEventType.TA_DIAGNOSTICS_UPDATE,
DIAGNOSTIC_INFORMATION_UPDATE_TRANSITION)
.addTransition(TaskAttemptStateInternal.SUCCESS_CONTAINER_CLEANUP,
EnumSet.of(TaskAttemptStateInternal.SUCCEEDED,
TaskAttemptStateInternal.KILLED),
TaskAttemptEventType.TA_KILL,
new KilledAfterSuccessTransition())
// Ignore-able events
.addTransition(TaskAttemptStateInternal.SUCCESS_CONTAINER_CLEANUP,
TaskAttemptStateInternal.SUCCESS_CONTAINER_CLEANUP,
EnumSet.of(TaskAttemptEventType.TA_KILL,
EnumSet.of(
TaskAttemptEventType.TA_FAILMSG,
TaskAttemptEventType.TA_FAILMSG_BY_CLIENT,
TaskAttemptEventType.TA_TIMED_OUT,

30
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

@ -1870,6 +1870,36 @@ public class TestTaskAttempt{
createReduceTaskAttemptImplForTest(eventHandler, clock, jobConf);
}
@Test
public void testKillingTaskWhenContainerCleanup() {
MockEventHandler eventHandler = new MockEventHandler();
TaskAttemptImpl taImpl = createTaskAttemptImpl(eventHandler);
TaskId maptaskId = MRBuilderUtils.newTaskId(taImpl.getID().getTaskId()
.getJobId(), 1, TaskType.MAP);
TaskAttemptId mapTAId =
MRBuilderUtils.newTaskAttemptId(maptaskId, 0);
// move in two steps to the desired state (cannot get there directly)
taImpl.handle(new TaskAttemptEvent(taImpl.getID(),
TaskAttemptEventType.TA_DONE));
assertEquals("Task attempt's internal state is not " +
"SUCCESS_FINISHING_CONTAINER",
TaskAttemptStateInternal.SUCCESS_FINISHING_CONTAINER,
taImpl.getInternalState());
taImpl.handle(new TaskAttemptEvent(taImpl.getID(),
TaskAttemptEventType.TA_TIMED_OUT));
assertEquals("Task attempt's internal state is not " +
"SUCCESS_CONTAINER_CLEANUP",
TaskAttemptStateInternal.SUCCESS_CONTAINER_CLEANUP,
taImpl.getInternalState());
taImpl.handle(new TaskAttemptKillEvent(mapTAId, "", true));
assertEquals("Task attempt is not in KILLED state",
TaskAttemptState.KILLED,
taImpl.getState());
}
@Test
public void testTooManyFetchFailureWhileContainerCleanup() {
MockEventHandler eventHandler = new MockEventHandler();