Details
-
Bug
-
Resolution: Fixed
-
Major
-
6.0.0
-
Untriaged
-
No
Description
Currently a single thread is responsible for cleaning up all relevant contents from metadata bucket. With timers now being stored in metadata bucket, it would make sense to distribute the work across all Eventing nodes in the cluster. That way undeploy could wrap up sooner. Would be good to have a tunable flag to control the level parallelism.