[
https://issues.apache.org/jira/browse/YARN-4597?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15567053#comment-15567053
]
Jian He commented on YARN-4597:
-------------------------------
Glanced through the patch, few comments:
- Wondering why KillWhileExitingTransition is added: It will override the
exitcode of the container. I feel in this case we should preserve the previous
exitcode, because the container failed/succeeded first before the kill event.
Or maybe we don't even need this transition, as the behavior is that the kill
is ignored.
- NodeManager#containerScheduler, NMContext#containerScheduler variable is not
used, we can remove.
- I think it's fine to not expose the ContainerState#SCHEDULED state to the
user, this state is mostly internal-facing and transient. User may not care
which intermediate state the container is at.
{code}
case SCHEDULED:
return org.apache.hadoop.yarn.api.records.ContainerState.SCHEDULED;
{code}
> Add SCHEDULE to NM container lifecycle
> --------------------------------------
>
> Key: YARN-4597
> URL: https://issues.apache.org/jira/browse/YARN-4597
> Project: Hadoop YARN
> Issue Type: Bug
> Components: nodemanager
> Reporter: Chris Douglas
> Assignee: Arun Suresh
> Attachments: YARN-4597.001.patch, YARN-4597.002.patch
>
>
> Currently, the NM immediately launches containers after resource
> localization. Several features could be more cleanly implemented if the NM
> included a separate stage for reserving resources.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
