[
https://issues.apache.org/jira/browse/YARN-5638?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15485366#comment-15485366
]
Li Lu commented on YARN-5638:
-----------------------------
Having put some more thoughts, I think adding the attempt ID as one part of
identifying collectors may not be a good idea. Collectors belong to
applications but not attempts. So it is possible that we map multiple app
attempts to one collector (if the collector is ran in a separate container). It
is also possible to have several collectors within one application, though. For
example, a node got separated from other nodes, and all applications running on
it will get relaunched. Thus, I think it make sense to preserve the old <appId,
collector> mapping when facing users, but internally we need to treat the
corner case that multiple collectors (one active, others stale) are mapped to
the same application.
> Introduce a collector Id to uniquely identify collectors and their creation
> order
> ---------------------------------------------------------------------------------
>
> Key: YARN-5638
> URL: https://issues.apache.org/jira/browse/YARN-5638
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: timelineserver
> Reporter: Li Lu
> Assignee: Li Lu
>
> As discussed in YARN-3359, we need to further identify timeline collectors
> and their creation order for better service discovery and resource isolation.
> This JIRA proposes to use <app, attempt, timestamp> to accurately identify
> each timeline collector.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
