Higher CPU usage due to unused FQDN nodes

Not a question, but some information that might help others.

Observed behaviour was higher CPU usage, unknown when it started, beyond the 30 days of on box statistics. Investigation with top (and 1) showed it was the mcpd process jumping around the different management CPUs and raising usage to 100%.

This occurred on both the active and standby unit, pretty much ruling out it being due to traffic.

On advice of our F5 partner (thanks) we put the logging level for MCP to debug for a short time.

This resulted in many of the following log lines:

 

Dec  1 15:00:00 hostname err mcpd[7623]: 01070726:3: Node address (::) in partition (Common) may not reference monitor (/Common/hm_name)

 

It turned out this health monitor (hm_name) was active on a FQDN node which wasn’t used in any pool. Resulting in not looking up an IP for it it seems.

Once these unused FQDN nodes were deleted, the CPU usage returned to normal.