Mailing-List: contact yarn-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Date: Wed, 22 Feb 2017 23:27:44 +0000 (UTC)
From: "Daniel Templeton (JIRA)" <jira@apache.org>
To: yarn-issues@hadoop.apache.org
Message-ID: <JIRA.13044687.1487643486000.37105.1487806064167@Atlassian.JIRA>
In-Reply-To: <JIRA.13044687.1487643486000@Atlassian.JIRA>
References: <JIRA.13044687.1487643486000@Atlassian.JIRA> <JIRA.13044687.1487643486380@jira-lw-us.apache.org>
Subject: [jira] [Commented] (YARN-6210) FS: Node reservations can interfere
 with preemption
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Wed, 22 Feb 2017 23:27:48 -0000


    [ https://issues.apache.org/jira/browse/YARN-6210?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15879463#comment-15879463 ] 

Daniel Templeton commented on YARN-6210:
----------------------------------------

You're gonna love me for this one...

We now have:
{code}
    if (!starved) {
      lastTimeAtFairShare = now;
    }

    if (!starved ||
        now - lastTimeAtFairShare < fsQueue.getFairSharePreemptionTimeout()) {
      fairshareStarvation = Resources.none();
    } 
{code}

What about:

{code}
    if (!starved ||
        now - lastTimeAtFairShare < fsQueue.getFairSharePreemptionTimeout()) {
      if (!starved) {
        lastTimeAtFairShare = now;
      }

      fairshareStarvation = Resources.none();
    } 
{code}

?

> FS: Node reservations can interfere with preemption
> ---------------------------------------------------
>
>                 Key: YARN-6210
>                 URL: https://issues.apache.org/jira/browse/YARN-6210
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: fairscheduler
>    Affects Versions: 2.9.0
>            Reporter: Karthik Kambatla
>            Assignee: Karthik Kambatla
>         Attachments: YARN-6210.1.patch, YARN-6210.2.patch, YARN-6210.3.patch, YARN-6210.4.patch
>
>
> Today, on a saturated cluster, apps with pending demand reserve nodes. A new app might not be able to preempt resources because these nodes are already reserved. This can be reproduced by the example in YARN-6151. 
> Since node reservations are to prevent starvation of apps requesting large containers, triggering these reservations only on starved applications would avoid this situation. 


--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

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