[
https://issues.apache.org/jira/browse/FLINK-21144?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17272607#comment-17272607
]
Xintong Song commented on FLINK-21144:
--------------------------------------
After taking a closer look, I found it not trivial to support scheduling
periodical executions in `MainThreadExecutor`.
Scheduling periodical executions can be supported through
`RpcService#getScheduledExecutor()`. However, exposing the `ScheduledExecutor`
to the `RpcEndpoint#MainThreadExecutor` requires invasive changes to the
current encapsulation.
For now, I think we may only implement `MainThreadExecutor#schedule(callable,
delay, unit)`, which should be enough to fix the `tryResetPodCreationCoolDown`
problem.
I've opened a PR. Please take a look. [~fly_in_gis]
> KubernetesResourceManagerDriver#tryResetPodCreationCoolDown causes fatal error
> ------------------------------------------------------------------------------
>
> Key: FLINK-21144
> URL: https://issues.apache.org/jira/browse/FLINK-21144
> Project: Flink
> Issue Type: Bug
> Components: Deployment / Kubernetes
> Affects Versions: 1.12.1
> Reporter: Yang Wang
> Assignee: Xintong Song
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.12.2
>
>
> {{KubernetesResourceManagerDriver#tryResetPodCreationCoolDown}} is calling a
> not implemented method {{RpcEndpoint.MainThreadExecutor#schedule(Callable<V>
> callable, long delay, TimeUnit unit)}}. This will cause a fatal error and
> make JobManager terminate exceptionally.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
