Cluster with routing enabled has one node with consistently high CPU usage


We have an 8-node cluster with routing enabled and we're seeing consistently high CPU usage on one node in particular (over 90%). The other 7 nodes flash high at times but don't remain or do so as often.

Is this most likely a hotspot issue?

Also, just out of curiosity, does adaptive replica selection apply when routing is enabled? I know it's turned on by default but not sure if it is utilized in routing as in the default non-routing shard-allocation. So if, with routing enabled, I have P1 (P = primary) on the highly utilized node with R1 (R = replica) and R2 on a couple other less busy nodes, would P1 be generally avoided? My guess is that it would and that this could cause fluctuations in CPU usage across the 3 nodes?


Yep. Is the index getting a lot of index/query requests?

Good question, let me see if I can find out.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.