I noticed that when our memsql cluster is sitting idle, there was still a decent amount of CPU usage. It looks like memsql-ops was the largest consumer of CPU usage which surprised me. When I shut off the agent on one of our leaf nodes, I could see a clear difference in CPU usage below:
These lines represent our 6 leaf nodes and the bottom line represents the leaf node not running memsql-ops. I’m curious if there any settings we can tweak with or changes we could make that could reduce whatever activity is consuming this CPU.
For what it’s worth, even when our cluster is not idle, the difference in CPU usage between running memsql-ops and not running memsql-ops is clear in the graphs: