Hi, team. Has any observed a similar experience with ActiveResourceManager allocating more resources than expected after JM restart.
This specifically for 1.18 where
Let’s say we launch Flink app which needs 30 workers. Cluster is resource constrained and app can only register 15. If i kill JM which forces job restart, new JM will request workers I noticed in
https://github.com/apache/flink/blob/release-1.18/flink-runtime/src/main/java/org/[…]flink/runtime/resourcemanager/active/ActiveResourceManager.java. The declare worker count will flip between 30 (which is the intended config), and 45 (config + active worker count). It then flip flops these requests for quite some time thus taking up resources in the cluster. Has anyone seen this before?