hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jian He (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-6591) Embed NodeId as part of ContainerId ?
Date Fri, 12 May 2017 21:35:04 GMT

    [ https://issues.apache.org/jira/browse/YARN-6591?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16008719#comment-16008719

Jian He commented on YARN-6591:

Yeah, it would need be an additional field.
- We may append the NodeId in the String so that old code ContainerId#fromString can still
parse the containerId string with NodeId, and new code can still parse the ContainerId String
without the String.
- We may  also add equivalent new APIs in NMClient to use the NodeId so that callers using
old API still works. 

- For existing code which starts using the NodeId - the logic would be fallback to the old
way if the NodeId doesn't exist.
-- For example, one other use case may be the LogCLI, currently it queries RM or ATS to get
the NodeId for the container. With this change, this part can be avoided and the logic would
be if NodeId doesn't exist, use the old way to get the NodeId.

> Embed NodeId as part of ContainerId ?
> -------------------------------------
>                 Key: YARN-6591
>                 URL: https://issues.apache.org/jira/browse/YARN-6591
>             Project: Hadoop YARN
>          Issue Type: Bug
>            Reporter: Jian He
> It could be useful to have NodeId as part of ContainerId.
> - NMClient API can be simplified, as caller code won't need to track the Node for the
container. e.g. stopContainer has to provide the NodeId by the caller.
> - Usability and debugability can be improved as won't need to find the node for the container.

This message was sent by Atlassian JIRA

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

View raw message