[ 
https://issues.apache.org/jira/browse/YARN-2113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15961235#comment-15961235
 ] 

Eric Payne commented on YARN-2113:
----------------------------------

Thanks [~sunilg] for the update. I am still reviewing the patch, but I have a 
couple of minor things to say right now:


TestProportionalCapacityPreemptionPolicyIntraQueueUserLimit.java:
- Several comments in this file say "... assume minimum user limit factor is 
...". This should say "minimum user limit percent". This is a small thing, but 
to avoid confusion, I think it's important to make the distinction between 
minimum user limit percent and user limit factor.
{code}
     * Consider 2 users in a queue, assume minimum user limit factor is 50%.
{code}
- In several tests, the the {{root}} queue does not have values that equal the 
sums of the subqueues. For example, in the code below, shouldn't used be 100 
and pending be 50? Also, the comment says {{b}} instead of {{a}} for the 
queuename.
{code}
    String queuesConfig =
        // guaranteed,max,used,pending,reserved
        "root(=[100 100 55 170 0]);" + // root
            "-a(=[100 100 100 30 0])"; // b
{code}
- Also, the {{appsConfig}} often doesn't add up to the correct used numbers in 
the {{queuesConfig}}


> Add cross-user preemption within CapacityScheduler's leaf-queue
> ---------------------------------------------------------------
>
>                 Key: YARN-2113
>                 URL: https://issues.apache.org/jira/browse/YARN-2113
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: scheduler
>            Reporter: Vinod Kumar Vavilapalli
>            Assignee: Sunil G
>         Attachments: 
> TestNoIntraQueuePreemptionIfBelowUserLimitAndDifferentPrioritiesWithExtraUsers.txt,
>  YARN-2113.0001.patch, YARN-2113.0002.patch, YARN-2113.0003.patch, 
> YARN-2113.0004.patch, YARN-2113.0005.patch, YARN-2113.0006.patch, 
> YARN-2113.v0.patch
>
>
> Preemption today only works across queues and moves around resources across 
> queues per demand and usage. We should also have user-level preemption within 
> a queue, to balance capacity across users in a predictable manner.



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

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

Reply via email to