[
https://issues.apache.org/jira/browse/YARN-5889?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15749453#comment-15749453
]
Jason Lowe commented on YARN-5889:
----------------------------------
I think I can get behind a proposal that preserves the FIFO/priority behavior
at least up through usage < MULP and then becomes fair once a user is beyond
MULP. We need to be careful that active users are still allowed to grow to
fill the queue (e.g.: a single user in the queue doesn't get capped at MULP).
Also note that a cached active user limit calculation will become out-of-date
more often than before, since resource-used-by-active-users changes a lot more
often than queue-capacity does. Essentially any resource change for an active
user invalidates the user limit cache for all active users. I'm not too
worried about that, just pointing out that it will get computed more often with
the new proposal.
> Improve user-limit calculation in capacity scheduler
> ----------------------------------------------------
>
> Key: YARN-5889
> URL: https://issues.apache.org/jira/browse/YARN-5889
> Project: Hadoop YARN
> Issue Type: Bug
> Components: capacity scheduler
> Reporter: Sunil G
> Assignee: Sunil G
> Attachments: YARN-5889.v0.patch, YARN-5889.v1.patch,
> YARN-5889.v2.patch
>
>
> Currently user-limit is computed during every heartbeat allocation cycle with
> a write lock. To improve performance, this tickets is focussing on moving
> user-limit calculation out of heartbeat allocation flow.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
