Gyula Fora created FLINK-34619: ---------------------------------- Summary: Do not wait for scaling completion in UPGRADE state with in-place scaling Key: FLINK-34619 URL: https://issues.apache.org/jira/browse/FLINK-34619 Project: Flink Issue Type: Improvement Components: Kubernetes Operator Affects Versions: kubernetes-operator-1.7.0 Reporter: Gyula Fora Assignee: Gyula Fora Fix For: kubernetes-operator-1.8.0
The operator currently puts the resource into upgrading state after triggering in-place scaling and keeps observing until the desired parallelism is reached before moving to deployed / stable. However this means that due to how the adaptive scheduler works this parallelism may never be reached and this is expected. We should simplify the logic to consider scaling "done" once the resource requirements have been set correctly and then leave the rest to the adaptive scheduler -- This message was sent by Atlassian Jira (v8.20.10#820010)