This message was deleted.

Check out this article

Supervisors are not there for compaction tasks. Supervisors are only there for real time ingestion afaik. From what I get is that you want to limit how much data source is compacted. Is there a slot bottleneck on the cluster?