[ https://issues.apache.org/jira/browse/FLINK-26538 ]
Nicholas Jiang deleted comment on FLINK-26538:
----------------------------------------
was (Author: nicholasjiang):
>From the user perspective, most of the users desire to only a single operation
>to restart deployment, the simpler the better. The user doesn't care the
>solution of multiple process step which including the job state convert with
>the state machine knowledge, just restart in one operation better.
> Ability to restart deployment w/o spec change
> ---------------------------------------------
>
> Key: FLINK-26538
> URL: https://issues.apache.org/jira/browse/FLINK-26538
> Project: Flink
> Issue Type: Sub-task
> Components: Kubernetes Operator
> Reporter: Thomas Weise
> Priority: Major
>
> Operator should allow restart of the Flink deployment w/o any other spec
> change. This provides the escape hatch for an operator to recover a
> deployment that has gone into a bad state (for whatever reason including
> memory leaks, hung JVM etc.) without direct access to the k8s cluster. This
> can be addressed by adding a restartNonce to the CRD.
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
