Does anyone have some good advice on converting a native auto compaction supervisor to the MSQ engine? I wanted to give it a go, caveat is each segmentGranularity in producing about 162B rows and 2.1T of segments. I’ve looked through the docs and tried adding as much to the taskContext that would seem good to do but CPU utilization seems very low on the middlemanagers while running

Is reducing compaction time your end goal here? In that case maxNumTasks could be one taskContext parameter that you try if not already done.

Yes that’s the primary motivation. I currently have maxNumTasks set to 1000 which is the max it could be set it from what I understand in the docs