Cores Pegged at 100%

I have a 16 node cluster with 32 cores per node. I’m noticing that a handful of cores are being pegged at 100% throughout the cluster. The cluster is only running the data service, and the CPU workload is very light. Due to the pegged cores I’m seeing constant coordinator swapping. “Haven’t heard from a higher priority node or a master, so I’m taking over.”

I’m currently running 4.1.0-5005 Community Edition (build-5005)

What could be the cause of the pegged cores?

That’s a pretty old release now. I vaguely recall a bug back in 4.x which would result in a worker thread busy-waiting. I recommend you upgrade to the current release (5.1.0 EE, 5.0.1 CE) and try again.

Is it possible to go from 4.1 CE to 5.0.1 CE? The upgrade matrix only mentions EE