The MOM start crashing. We found there are several CLOSE_WAIT connections and it is increasing. When the limit is reached at MOM, the MOM stop working. The only way to resolve this issue to recycle MOM. The issue occurs again when this condition is meet. This is occurring in both PROD and NON-PROD environments.
It seems like issue is cause by older agents (EPAgent and Java Agents) (some from APM release 9.6) that are reporting metrics to this APM cluster.
Release : 10.7.0
Component : Introscope
For old EPAgent, stop the process.
For old WAS Agent, try disable Agent for now. If it is not feasible to do that, block the IP from MOM side
isengard.server.banned.client.ipaddresses=<',' separated IPs>