[
https://issues.apache.org/jira/browse/YARN-7848?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16798387#comment-16798387
]
Eric Badger commented on YARN-7848:
-----------------------------------
I don't like this idea. This is the Nodemanager completely taking over docker.
The NM should use docker, but I don't think it should be assumed that it is the
only thing that can use docker. To me, this seems like something that should be
handled at an ops level in a cron job, if they want to make sure all images are
gone. I'm ok with having a more generalized case of pruning images
periodically, but I'm not comfortable with the NM pruning containers that it
didn't start.
Additionally, how would this work with debug-delay? If I want my image to stick
around for awhile (or indefinitely) so that I can debug them, how will that
co-exist with this periodic pruning?
cc [[email protected]]
> Force removal of docker containers that do not get removed on first try
> -----------------------------------------------------------------------
>
> Key: YARN-7848
> URL: https://issues.apache.org/jira/browse/YARN-7848
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Eric Badger
> Assignee: Zhaohui Xin
> Priority: Major
> Labels: Docker
>
> After the addition of YARN-5366, containers will get removed after a certain
> debug delay. However, this is a one-time effort. If the removal fails for
> whatever reason, the container will persist. We need to add a mechanism for a
> forced removal of those containers.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
