[jira] [Updated] (YARN-7401) Reduce lock contention in ClusterNodeTracker#getClusterResource()

["Daniel Templeton (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/YARN-7401?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Daniel Templeton updated YARN-7401:
-----------------------------------
    Attachment: YARN-7401.001.patch

To get rid of the lock in {{getClusterCapacity()}}, I moved the copy from {{clusterCapacity}} to where {{clusterCapacity}} is modified.  Then by making {{staleClusterCapacity}} volatile, I can remove the lock.  Tests shows a significant fair scheduler performance increase with this patch.  I suspect capacity scheduler will also benefit.

> Reduce lock contention in ClusterNodeTracker#getClusterResource()
> -----------------------------------------------------------------
>
>                 Key: YARN-7401
>                 URL: https://issues.apache.org/jira/browse/YARN-7401
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: resourcemanager
>    Affects Versions: 3.1.0
>            Reporter: Daniel Templeton
>            Assignee: Daniel Templeton
>         Attachments: YARN-7401.001.patch
>
>
> Profiling the code shows massive latency in {{ClusterNodeTracker.getClusterResource()}} on getting the lock.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

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