You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Bibin Chundatt (Jira)" <ji...@apache.org> on 2019/09/27 05:47:00 UTC

[jira] [Comment Edited] (YARN-9858) Optimize RMContext getExclusiveEnforcedPartitions

    [ https://issues.apache.org/jira/browse/YARN-9858?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16939136#comment-16939136 ] 

Bibin Chundatt edited comment on YARN-9858 at 9/27/19 5:46 AM:
---------------------------------------------------------------

[~jhung]

Patch could causeĀ  *exclusiveEnforcedPartitions* getting set multiple times incase of concurrent execution.
 Its a possibility since its invoked by multiple handler.

Alternative could be to set the *exclusiveEnforcedPartitions* after the creation of RMContext at
 # Resourcemanager#serviceInit
 # Resourcemanager#resetRMContext

All activeservices would be in NEW STATE when set it too.  Thoughts?


was (Author: bibinchundatt):
[~jhung]

Patch could causeĀ  *exclusiveEnforcedPartitions* getting set multiple times incase of concurrent execution.
 Its a possibility since its invoked by multiple handler.

Alternative could be to set the *exclusiveEnforcedPartitions* after the creation of RMContext at
 # Resourcemanager#serviceInit
 # Resourcemanager#resetRMContext

All the activeservices would be in stopped when we set it too.. Thoughts?

> Optimize RMContext getExclusiveEnforcedPartitions 
> --------------------------------------------------
>
>                 Key: YARN-9858
>                 URL: https://issues.apache.org/jira/browse/YARN-9858
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Jonathan Hung
>            Assignee: Jonathan Hung
>            Priority: Major
>              Labels: release-blocker
>         Attachments: YARN-9858.001.patch
>
>
> Follow-up from YARN-9730. RMContextImpl#getExclusiveEnforcedPartitions is a hot code path, need to optimize it .
> Since AMS allocate invoked by multiple handlers locking on conf will occur
> {code}
> java.lang.Thread.State: BLOCKED (on object monitor)
>  at org.apache.hadoop.conf.Configuration.getProps(Configuration.java:2841)
>  - waiting to lock <0x00007f1f8107c748> (a org.apache.hadoop.yarn.conf.YarnConfiguration)
>  at org.apache.hadoop.conf.Configuration.get(Configuration.java:1214)
>  at org.apache.hadoop.conf.Configuration.getTrimmed(Configuration.java:1268)
> {code}



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

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