[
https://issues.apache.org/jira/browse/YARN-3390?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14509962#comment-14509962
]
Zhijie Shen commented on YARN-3390:
-----------------------------------
Thanks for the comments. I've addressed Sangjin and Li's comments except:
bq. maybe we'd like to mark it as unstable?
It's not the API for the users, hence it's okay to leave it unannotated.
bq. In TimelineCollectorWebService, why we're removing the utility function
getCollector?
After the refactoring, we don't need to convert appId to string. It's not
necessary to wrap a single statement in a method.
In addition, I changed to use hook in TimelineCollectorManager, but postRemove
is called before stopping the collector, because once the collector is stopped,
the hook may not be able to do something with the stopped collector.
Moreover, I moved RMApp.stopTimelineCollector into FinalTransition. Suppose the
collector only collects application lifecycle events, it doesn't need to stay
after the app is finished. We can adjust it later if we find the collector
needs to stay after the app is finished.
> Reuse TimelineCollectorManager for RM
> -------------------------------------
>
> Key: YARN-3390
> URL: https://issues.apache.org/jira/browse/YARN-3390
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: timelineserver
> Reporter: Zhijie Shen
> Assignee: Zhijie Shen
> Attachments: YARN-3390.1.patch, YARN-3390.2.patch
>
>
> RMTimelineCollector should have the context info of each app whose entity
> has been put
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
