Hello Robin,
We had the same problem. We upgraded from 2.0.p27 on CentOS 7 to 2.1.p13 on CentOS 7 and then restore configuration to RockyLinux 9.1. Our CPU problem was identical to the OP’s. Im attaching below a screenshot of CPU usage, as u can see on the left there are statistics from CentOS and then a high spike from Rocky.
Then, as a workaround, we wrote a script to reload the CMC service every time it reached 90%.
Yesterday we disabled SNMP on each of our 6 APC UPS and it worked like a charm.
Here are VMware statistics:
I will let you know in a couple of days which device is sending the broken SNMP request.
In addition, I want to let you know that from the htop it looked like the CMC process reserved 99%, but each part of the process generally used 0% and then when, for example when data collect from agent, one part of the CMC process wake up for about 10% in couple seconds and then disappeared in queue to take another action. I’m sorry, but I don’t have a screenshot of it.