Hi Ferenc, Thanks for pointing this out! After confirmation with others, I think you are right.
It's fair enough to revert the PR that remove per-job mode in 2.0-preview. But we need to make sure we don't include per-job stuff in 2.0.0. Best regards, Weijie Ferenc Csaky <ferenc.cs...@pm.me.invalid> 于2024年10月9日周三 21:55写道： > Hello devs, > > I just noticed there is an already merged change [1] that removes per-job > mode from the > codebase. Regarding that, I would like to mention FLINK-26000 [2] that > exists for the same > for quite a while, and the reason why we kept the code until now is that, > Application Mode > does not support SQL job submission. AFAIK that is still the case. > FLIP-316 [3] aimed to fill > this gap utilizing the SQL Gateway, although it is stale for quite a while. > > So if I am not missing something, as it is right now Flink 2.0 will lose > the ability of deploying > SQL jobs to their dedicated cluster on YARN. Is this something the > community considered? > I understand that there is more focus around K8s nowadays, but we for > example have a > platform that operates on YARN, and most of our customers use that, where > per-job mode is > critical for resource isolation. > > Looking forward for your thoughts! > > Best, > Ferenc > > [1] https://issues.apache.org/jira/browse/FLINK-36310 > [2] https://issues.apache.org/jira/browse/FLINK-26000 > [3] > https://cwiki.apache.org/confluence/display/FLINK/FLIP-316%3A+Support+application+mode+for+SQL+Gateway
