[
https://issues.apache.org/jira/browse/TEZ-3097?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15611534#comment-15611534
]
Harish Jaiprakash commented on TEZ-3097:
----------------------------------------
I was able to approximately replicate the issue: (Not the concurrent
modification but get events after dag finished).
* delay enqueueAndScheduleNextEvent(new
VertexManagerEventOnVertexStarted(completions)) in
VertexManager.onVertexStarted by 20-30 ms in a different thread.
* This is to simulate a case where the executerService would not have invoked
the callback for onVertexStarted, but the task complete events for the vertex
has been handled and dag is finished.
* Now we have a finished dag, and suddenly
ShuffleVertexManagerBase.OnVertexStarted is scheduled by the VertexManager and
which caused new tasks to be scheduled and events sent to HistoryEventHandler.
Followup:
* Is this a bug? Should we handle case where dag can be finished before the
above schedule is done. (No test fix required in this case).
* Is there a way to fire the task finished events only after Vertex has fully
started. More like, fire task finished only after task start is called? Only
way I can think of is to wait for task started events to arrive in
MockHistoryEventHandler (make it thread safe too).
* These extra events do not affect the outcome, use a concurrent queue instead
of the list to make MockHistoryEventHandler thread safe.
> Flaky test: TestCommit.testDAGCommitStartedEventFail_OnDAGSuccess
> -----------------------------------------------------------------
>
> Key: TEZ-3097
> URL: https://issues.apache.org/jira/browse/TEZ-3097
> Project: Apache Tez
> Issue Type: Bug
> Affects Versions: 0.8.3
> Reporter: Jeff Zhang
>
> {noformat}
> testDAGCommitStartedEventFail_OnDAGSuccess(org.apache.tez.dag.app.dag.impl.TestCommit)
> Time elapsed: 0.072 sec <<< ERROR!
> java.util.ConcurrentModificationException: null
> at java.util.ArrayList$Itr.checkForComodification(ArrayList.java:859)
> at java.util.ArrayList$Itr.next(ArrayList.java:831)
> at
> org.apache.tez.dag.app.dag.impl.TestCommit$MockHistoryEventHandler.verifyVertexCommitStartedEvent(TestCommit.java:2033)
> at
> org.apache.tez.dag.app.dag.impl.TestCommit.testDAGCommitStartedEventFail_OnDAGSuccess(TestCommit.java:1880)
> {noformat}
> https://builds.apache.org/job/Tez-Build-Hadoop-2.4/245/console
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
