[
https://issues.apache.org/jira/browse/FLINK-36577?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18005166#comment-18005166
]
Mate Czagany commented on FLINK-36577:
--------------------------------------
That's an interesting idea indeed, and a bit more exciting than the scope of
this ticket too.
To know the operator UUIDs, the Kubernetes Operator would have to have access
to the JAR somehow. Or we have a Flink JobManager running with version v1.1,
and with a new API endpoint, we could supply the URI of job v1.2 and a
savepoint path, and the JobManager would do the actual checks. Flink Kubernetes
Operator would just call this endpoint as it calls any other endpoint.
Definitely a much bigger task, do you see this as a useful feature? Do you
think it would result in a faster feedback loop than currently available? For
example, users just launching the job in their dev environment and checking the
JM logs if it fails.
> Add compatibility check for FlinkStateSnapshot CRD
> --------------------------------------------------
>
> Key: FLINK-36577
> URL: https://issues.apache.org/jira/browse/FLINK-36577
> Project: Flink
> Issue Type: Improvement
> Components: Kubernetes Operator
> Reporter: Mate Czagany
> Assignee: Mate Czagany
> Priority: Major
> Labels: pull-request-available
> Fix For: kubernetes-operator-1.13.0
>
>
> There are already two CRD compatibility checks in the
> flink-kubernetes-operator-api POM (for FlinkDeployment and FlinkSessionJob),
> one should also be added for FlinkStateSnapshot.
>
> Requirement for this is to have the 1.10.0 version released.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
