[
https://issues.apache.org/jira/browse/FLINK-21667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17298496#comment-17298496
]
Xintong Song commented on FLINK-21667:
--------------------------------------
Thanks [~trohrmann], I think that's a really good idea.
The lifecycle management and leadership management are coupled for the
{{ResourceManager}}. It should simplify things a lot if we only start the
{{ResourceManager}} when leadership is obtained, so that the
{{ResourceManager}} and its components can always assume having the leadership.
I'll take a closer look into this.
BTW, I assume what you meant is {{Dispatcher}} and {{JobMaster}}? We don't have
leader election for {{TaskExecutor}}.
> Standby RM might remove resources from Kubernetes
> -------------------------------------------------
>
> Key: FLINK-21667
> URL: https://issues.apache.org/jira/browse/FLINK-21667
> Project: Flink
> Issue Type: Bug
> Components: Deployment / Kubernetes
> Affects Versions: 1.12.2
> Reporter: Xintong Song
> Priority: Major
> Fix For: 1.13.0, 1.12.3
>
>
> Currently, on initialization {{KubernetesResourceManagerDriver}} starts a
> watch for receiving pod events. It could happen that it starts to receive
> events before obtaining leadership. Consequently, a standby RM may remove
> terminated pods from Kubernetes during handling the events.
> This is not very damaging atm, since the removed pods are already terminated
> anyway. However, it would still be good for a standby RM to strictly
> following the contract and make no modifications before obtaining leadership.
> We might consider to postpone starting of the watch to when the leadership is
> granted.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
