[
https://issues.apache.org/jira/browse/FLINK-6183?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15940910#comment-15940910
]
ASF GitHub Bot commented on FLINK-6183:
---------------------------------------
GitHub user zentol opened a pull request:
https://github.com/apache/flink/pull/3610
[FLINK-6183]/[FLINK-6184] Prevent some NPE and unclosed metric groups
This PR fixes 2 issues:
1) It prevents some NPEs in the buffer metrics by instantiating them after
the task has been registered in the NetworkEnvironment.
2) It prevents some cases where the TaskMetricGroup would never be closed.
These cases include an early exit in `Task#run()` and when 2) tasks with an
identical ExecutionAttemptID are run on the same TM.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/zentol/flink 6183_6184_metric_task
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/flink/pull/3610.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #3610
----
----
> TaskMetricGroup may not be cleanup when Task.run() is never called or exits
> early
> ---------------------------------------------------------------------------------
>
> Key: FLINK-6183
> URL: https://issues.apache.org/jira/browse/FLINK-6183
> Project: Flink
> Issue Type: Bug
> Components: Metrics
> Affects Versions: 1.2.0, 1.3.0
> Reporter: Chesnay Schepler
> Assignee: Chesnay Schepler
> Priority: Blocker
>
> The TaskMetricGroup is created when a Task is created. It is cleaned up at
> the end of Task.run() in the finally block. If however run() is never called
> due some failure between the creation and the call to run the metric group is
> never closed. This also means that the JobMetricGroup is never closed.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
