[
https://issues.apache.org/jira/browse/FLINK-11343?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16775172#comment-16775172
]
TisonKun commented on FLINK-11343:
----------------------------------
[~dawidwys] I think I lost it and it seems that this scenario happens rarely.
The problematic order is theoretically possible as described at the
corresponding pull request.
1. {{Task#run}} set task state as FINISHED
2. Test case get the FINISHED future, call {{taskExecutor#shutdown}}, so the
MemoryManager shutdown.
3. In {{finally}} block of {{Task#run}}, call
{{MemoryManager#releaseAll(owner)}}
4. cause {{IllegalStateException}}
> TaskExecutorTest is unstable on travis
> --------------------------------------
>
> Key: FLINK-11343
> URL: https://issues.apache.org/jira/browse/FLINK-11343
> Project: Flink
> Issue Type: Bug
> Components: Tests
> Affects Versions: 1.8.0
> Reporter: TisonKun
> Assignee: TisonKun
> Priority: Major
> Labels: pull-request-available, test-stability
> Fix For: 1.8.0
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> {code:java}
> org.apache.flink.runtime.util.TestingFatalErrorHandler$TestingException:
> java.lang.IllegalStateException: Memory manager has been shut down.
> at
> org.apache.flink.runtime.util.TestingFatalErrorHandler.rethrowError(TestingFatalErrorHandler.java:51)
> at
> org.apache.flink.runtime.taskexecutor.TaskExecutorTest.teardown(TaskExecutorTest.java:223)
> Caused by: java.lang.IllegalStateException: Memory manager has been shut down.
> at
> org.apache.flink.runtime.memory.MemoryManager.releaseAll(MemoryManager.java:480)
> at org.apache.flink.runtime.taskmanager.Task.run(Task.java:821)
> at java.lang.Thread.run(Thread.java:748)
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
