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 "Vrushali C (JIRA)" <ji...@apache.org> on 2018/07/06 19:01:00 UTC

[jira] [Commented] (YARN-6767) Timeline client won't be able to write when TimelineCollector is not up yet, or NM is down

    [ https://issues.apache.org/jira/browse/YARN-6767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16535235#comment-16535235 ] 

Vrushali C commented on YARN-6767:
----------------------------------

We need to handle collector fault tolerance, that is the case of collector going down after it had come up. 

But the case of collector not being up itself in the first place needs to be handled by the client framework. 

> Timeline client won't be able to write when TimelineCollector is not up yet, or NM is down
> ------------------------------------------------------------------------------------------
>
>                 Key: YARN-6767
>                 URL: https://issues.apache.org/jira/browse/YARN-6767
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineclient
>    Affects Versions: 3.0.0-alpha4
>            Reporter: Haibo Chen
>            Priority: Major
>
> As discussed in the call, when an application first starts to run, its corresponding TimelineCollector instance may not be up yet, or if the TimelineCollector goes down when node manager dies (TimelineCollector now runs as part of NM auxiliary services), the timeline client
> will not able to write entities. We need to address or mitigate the issue if possible, or at least call it out.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

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