[
https://issues.apache.org/jira/browse/FLINK-22594?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17340766#comment-17340766
]
Aitozi commented on FLINK-22594:
--------------------------------
If container oom killed due to run out of memory, the container status will
exposed the OOMKilled msg[1], it can be catch in k8s watcher. It can be a guide
for user to correct the memory of worker. we can do the improvement by
1) add a handler to catch the related msg and expose to metric
2) support auto expand the memory of next requested pod of some ratio to let
the case self-handling
[1]:
https://kubernetes.io/docs/tasks/configure-pod-container/assign-memory-resource/#exceed-a-container-s-memory-limit
> Report oom killed pod in k8s resource manager
> ---------------------------------------------
>
> Key: FLINK-22594
> URL: https://issues.apache.org/jira/browse/FLINK-22594
> Project: Flink
> Issue Type: Improvement
> Components: Deployment / Kubernetes
> Reporter: Aitozi
> Priority: Minor
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
