[
https://issues.apache.org/jira/browse/FLINK-3266?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15134386#comment-15134386
]
Stephan Ewen commented on FLINK-3266:
-------------------------------------
Can you verify they stay alive after a single execution with the
LocalEnvironment? (Ideally get a thread sump after one execution, because this
thread dump already has some thread from a cluster being started).
I can think of two possible causes from the top of my head
- The local mini-cluster does not shut down Akka properly
- The Scala Global Execution Context is somehow leaky
> LocalFlinkMiniCluster leaks resources when multiple jobs are submitted
> ----------------------------------------------------------------------
>
> Key: FLINK-3266
> URL: https://issues.apache.org/jira/browse/FLINK-3266
> Project: Flink
> Issue Type: Bug
> Components: Local Runtime
> Affects Versions: 1.0.0
> Reporter: Gabor Gevay
> Priority: Minor
>
> After a job submitted to a LocalEnvironment finishes, some threads are not
> stopped, and are stuck in waiting forever.
> You can observe this, if you enclose the body of the main function of the
> WordCount example with a loop that executes 100 times, and monitor the thread
> count (with VisualVM for example).
> (The problem only happens if I use a mini cluster. If I use start-local.sh
> and submit jobs to it, then there is no leak.)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
