[jira] [Commented] (YARN-3491) PublicLocalizer#addResource is too slow.

["Jason Lowe (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/YARN-3491?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14498092#comment-14498092 ] 

Jason Lowe commented on YARN-3491:
----------------------------------

Storing asynchronously is going to be a bit dangerous -- we do not want to create a situation where a resource has started localizing but we haven't recorded the fact that we started it.  Theoretically we could end up doing a recovery where we leak a resource or fail to realize a localization started but did not complete and we need to clean it up.

I think it's best at this point to have some hard evidence from a profiler or targeted log statements around the suspected code where all the time is being spent in the NM rather than guessing.

> PublicLocalizer#addResource is too slow.
> ----------------------------------------
>
>                 Key: YARN-3491
>                 URL: https://issues.apache.org/jira/browse/YARN-3491
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: nodemanager
>    Affects Versions: 2.7.0
>            Reporter: zhihai xu
>            Assignee: zhihai xu
>            Priority: Critical
>
> Improve the public resource localization to do both FSDownload submission to the thread pool and completed localization handling in one thread (PublicLocalizer).
> Currently FSDownload submission to the thread pool is done in PublicLocalizer#addResource which is running in Dispatcher thread and completed localization handling is done in PublicLocalizer#run which is running in PublicLocalizer thread.
> Because PublicLocalizer#addResource is time consuming, the thread pool can't be fully utilized. Instead of doing public resource localization in parallel(multithreading), public resource localization is serialized most of the time.
> Also there are two more benefits with this change:
> 1. The Dispatcher thread won't be blocked by PublicLocalizer#addResource . Dispatcher thread handles most of time critical events at Node manager.
> 2. don't need synchronization on HashMap (pending).
> Because pending will be only accessed in PublicLocalizer thread.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)