hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sunil G (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-5889) Improve user-limit calculation in capacity scheduler
Date Mon, 06 Feb 2017 07:35:42 GMT

    [ https://issues.apache.org/jira/browse/YARN-5889?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15853626#comment-15853626
] 

Sunil G commented on YARN-5889:
-------------------------------

HI [~eepayne]. 
Sorry for troubling you again. I ran many similar test cases in my local cluster. I could
not get the case mentioned by you, even though i tested for many time today and yesterday.
So i thought I might be missing something. 

I will try share my test case results here:
- I used {{YARN-5889.0010.patch}} patch for my tests.
- One node cluster (8GB).
- I used sleep job to make sure i have a control on test time and on number of containers.
- AM and other containers were using 2G containers always.


*Test Scenario* (For 3 test cases, same scenario were used, only change is in configuration)
1) Ran 4 containers including AM. Pending RR is 2Gb for one more container
2) Killed all non-AM containers 3 times (totally 9 containers were killed using FORCEFUL_SHUTDOWN
signal)
3) Killed AM containers once and job re-ran.

+Scenario 1:+ (One Queue 100%)
MULP = 100, ULFactor = 1

Output:
Application retook whole cluster always when one/more container were killed.

+Scenario 2:+ (One Queue 100%)
MULP = 50, ULFactor = 1


Output:
Application retook whole cluster always when one/more container were killed. UL was 50%, still
it got whole cluster.


+Scenario 3:+ (Two Queues 50% each)
MULP = 50, ULFactor = 2

Output:
Application retook whole cluster always when one/more container were killed. UL was 50%, still
it got whole cluster as UL factor was 2. (If its 1, then only 6Gb could be used).

Could you please help to take a look and share me what I might have missed from your test
scenario. I ll continue test some more scenarios and will keep posted.

> 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-5889.0001.patch, YARN-5889.0001.suggested.patchnotes, YARN-5889.0002.patch,
YARN-5889.0003.patch, YARN-5889.0004.patch, YARN-5889.0005.patch, YARN-5889.0006.patch, YARN-5889.0007.patch,
YARN-5889.0008.patch, YARN-5889.0009.patch, YARN-5889.0010.patch, 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
(v6.3.15#6346)

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


Mime
View raw message