Hi everone, I am working with the Apache Flink Operator but I have an issue with the getParallelism method, which always override my parallelism setting in yaml configuration. Let's say I have a Flink application running with Application Mode and Standalone which has 3 operator tasks and they run in their own slot sharing groups, e.g. operator A runs in SlotSharingGroup A, operator B runs in SlotSharingGroup B,...Its parallelism is 2, so the Flink job needs at least 6 task slots. If

taskmanager.numberOfTaskSlots

is 2, the

taskManager.replicas

should be at least 3. However, with the implemenation of the getParallelism method, the parallelism.default will be *6 = `taskmanager.numberOfTaskSlots * taskManager.replicas`* , so it overrides my parallelism setting of 2. Therefore, I think the logic of the getParallelism method, does not make sense to me. Do I miss anything? Thanks

Bumping this up for visibility. Has anyone seen this and understood why the implementation is as such?

It’s impossible to know what slot sharing groups the job defines