[
https://issues.apache.org/jira/browse/YARN-9788?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956957#comment-16956957
]
Peter Bacsko commented on YARN-9788:
------------------------------------
Thanks for the patch [~Prabhu Joseph]. I think the patch looks good.
Really just one nitpick: the test name {{testParallelUpdates}} might be
slightly misleading because things are not really happening in parallel. It's
more like making sure that updates are not lost. So a better name would be sth
like {{testMultipleUpdates}} or {{testMultipleUpdatesNotLost}}, etc.
Otherwise +1 non-binding.
> Queue Management API - does not support parallel updates
> --------------------------------------------------------
>
> Key: YARN-9788
> URL: https://issues.apache.org/jira/browse/YARN-9788
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: capacity scheduler
> Affects Versions: 3.3.0
> Reporter: Prabhu Joseph
> Assignee: Prabhu Joseph
> Priority: Major
> Attachments: YARN-9788-001.patch, YARN-9788-002.patch,
> YARN-9788-003.patch, YARN-9788-004.patch, YARN-9788-005.patch,
> YARN-9788-006.patch, YARN-9788-007.patch
>
>
> Queue Management API - does not support parallel updates. When there are two
> parallel schedule conf updates (logAndApplyMutation), the first update is
> overwritten by the second one.
> Currently the logAndApplyMutation creates LogMutation and stores it in a
> variable pendingMutation. This way at any given time there will be only one
> LogMutation. And so the two parallel logAndApplyMutation will override the
> pendingMutation and the later one only will be present.
> The fix is to return LogMutation object by logAndApplyMutation which can be
> passed during confirmMutation. This fixes the parallel updates.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
