[
https://issues.apache.org/jira/browse/YARN-10996?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17437181#comment-17437181
]
Andras Gyori commented on YARN-10996:
-------------------------------------
It is important to check each location individually, as laborously replacing
getUser with getUserAndAddIfAbsent could introduce a memory leak as:
* users are only removed when an application attempt is removed belonging to
the user and the user has no more application left
* if we reinsert a user without an application, that user will never be
removed again, thus introducing a leak
> Fix race condition of User object acquisitions
> ----------------------------------------------
>
> Key: YARN-10996
> URL: https://issues.apache.org/jira/browse/YARN-10996
> Project: Hadoop YARN
> Issue Type: Bug
> Reporter: Andras Gyori
> Assignee: Andras Gyori
> Priority: Major
>
> UserManager is highly susceptible of removing users that are queried later.
> This race condition produces NPE similar to YARN-10934. A non-exhaustive list
> of these locations are:
> - LeafQueue.getTotalPendingResourcesConsideringUserLimit
> - UsersManager.computeUserLimit
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
