[
https://issues.apache.org/jira/browse/FLINK-32867?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17754646#comment-17754646
]
haiqingchen commented on FLINK-32867:
-------------------------------------
Thanks, I got the idea about savepoint upgradeMode. But if I want a full job
restart with a predefined checkpoint path, the only option is to delete the
deployment, wait until the deployment deleted completely and recreate the
deployment with initial savepoint path. I'm wondering if we could introduce
another upgrade mode option for this kind of situation: If I update
restartNonce with a predefined checkpoint path, the job will restart from the
checkpoint path.
> Suspending job and triggering savepoint when only updating restartNonce
> -----------------------------------------------------------------------
>
> Key: FLINK-32867
> URL: https://issues.apache.org/jira/browse/FLINK-32867
> Project: Flink
> Issue Type: Improvement
> Components: Kubernetes Operator
> Affects Versions: kubernetes-operator-1.4.0
> Reporter: haiqingchen
> Priority: Minor
>
> When I tried to restart Flink Deployment without any configuration change,
> I updated restartNonce to current time mills.
> However, kubernetes operator suspend the Flink Deployment and trigger a
> savepoint but we didn't update savepointTriggerNonce column.
> It's not what we expected as we thought if we didn't update
> savepointTriggerNonce it would never trigger savepoint.
> Is there any possibility to add a precheck(if savepointTriggerNonce is
> updated) when canceling the job?
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
