Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Fri, 16 Dec 2016 19:29:58 +0000 (UTC)
From: "Wangda Tan (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.13021186.1479315855000.542142.1481916598450@Atlassian.JIRA>
In-Reply-To: <JIRA.13021186.1479315855000@Atlassian.JIRA>
References: <JIRA.13021186.1479315855000@Atlassian.JIRA> <JIRA.13021186.1479315855295@arcas>
Subject: [jira] [Commented] (YARN-5889) Improve user-limit calculation in
 capacity scheduler
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Fri, 16 Dec 2016 19:30:05 -0000


    [ https://issues.apache.org/jira/browse/YARN-5889?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15755304#comment-15755304 ] 

Wangda Tan commented on YARN-5889:
----------------------------------

bq. For preemption calculation, one of the main problem could have been about the free resources
I think so, but I think the *minimum* preemption threshold is we should not preempt res from user if used of the user < queue-capacity * MULP. The higher threshold is, the safer of the preemption. 

bq. I think in a busier and short-living app's cluster, we may recalculate more.
I agree with [~eepayne], we should be able to make preemption / allocation UL calculation independent (or at least it's better to not allocation UL has dependency on preemption UL). [~sunilg] could you add some details here?

bq. On this note, could I update a patch with approach mentioned above.
Please go ahead when you get chance :). We should generally agree the approach, there are some debatable details, but starting prototype can help us understand the scope.

Thoughts?


> 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-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org