[
https://issues.apache.org/jira/browse/FLINK-29872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17628516#comment-17628516
]
Gyula Fora commented on FLINK-29872:
------------------------------------
With components such as the Kubernetes Operator these compatibilty questions
become more important. The current operator needs to talk to multiple Flink
versions at the same time to perform basic cluster operations (ideally using
the latest supported Flink client)
It is trivial to fix these incompatibilities on Flink rest side and that allows
us to use the operator without having to add copies of fixed classes to shadow
what comes from flink.
We simply should not introduce required fields in the rest api. It's easy not
to do it with some attention and I don't see a reason why we would anyway.
> JobDetailsInfo Rest endpoint breaking change in 1.16
> ----------------------------------------------------
>
> Key: FLINK-29872
> URL: https://issues.apache.org/jira/browse/FLINK-29872
> Project: Flink
> Issue Type: Bug
> Components: Runtime / REST
> Affects Versions: 1.16.0
> Reporter: Gyula Fora
> Priority: Major
>
> Flink 1.16 introduces a breaking change to the JobDetailsInfo endpoints by
> adding 3 required fields to:
> [https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/rest/messages/job/metrics/IOMetricsInfo.java]
> accumulatedBackpressured, accumulatedIdle, accumulatedBusy fields should not
> be primitive (which makes them required) but instead nullable.
> This would allow the 1.16 restclient to read the jobdetails of previous
> cluster versions
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
