[
https://issues.apache.org/jira/browse/YARN-4924?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15227703#comment-15227703
]
sandflee commented on YARN-4924:
--------------------------------
In YARN-4051, we also had containers leak from NEW to DONE transition while
recovering. The difference is in YARN-4051 NM received FINISH_APP event after
NM register while containers are not recovered.
As in YARN-4051, pending kill event is not suggested , may be we could delay
the sending of FINISH_APP event after containers are recovered. If this is
acceptable, I could update the patch in YARN-4051 to fix this.
> NM recovery race can lead to container not cleaned up
> -----------------------------------------------------
>
> Key: YARN-4924
> URL: https://issues.apache.org/jira/browse/YARN-4924
> Project: Hadoop YARN
> Issue Type: Bug
> Components: nodemanager
> Affects Versions: 3.0.0, 2.7.2
> Reporter: Nathan Roberts
>
> It's probably a small window but we observed a case where the NM crashed and
> then a container was not properly cleaned up during recovery.
> I will add details in first comment.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
