Hi All,
I have an issue between nexus 77k and prime infrastructure (PI) where ssh sessions at the PI end are kept in ESTABLISHED state and on the Nexus they appear as tcp sockets (they do not appear as vty lines). Nexus 77k are running multi-context and it occurs across admin and VDCs too.
As the N77k is not releasing the socket, the session stays up in Prime and eventually exhausts its 100 ssh thread pool. When this occurs, config archives and inventory is not able to function the server is so busy it is almost unusable.
Have seen the bug:
CSCue74597 (Stale ssh sessisons on N7K)
It mentions a workaround is to disable and enable ssh, or a workaround is to restart PI services.
Neither option are of much use to me in my customer environment
I have been able to trace the sockets to specific pids in the nexus, as there is no command to clear sockets in NX-OS.
Within two weeks of restarting PI services, already 30 of 100 threads in the PI pool are exhausted.
How do I kill specific PID in NX-OS 6.2(12) ?
Thanks in advance,
Greg.
