We created the state change log in
KAFKA-513 which creates a well formatted log of the lifetime of all state change requests in a Kafka cluster. While troubleshooting KAFKA-790, I looked into server.log, controller.log and state-change.log and realized we can improve the controller logging -
1. StopReplica state change is missing from the state-change.log
2. Some state change log messages are repeated. These are likely mistakenly inside a for loop rather than outside
3. Some controller logging that should be in the controller.log is still in server.log. For example, everything from ReplicaManager. This is useful for troubleshooting but confusing to be inside server.log.