Hey guys,

I have four zLinux SLES 11.3 servers which are monitored through net-snmp via Solarwinds. Recently when a manual resource listing (poll) is requested from the Solarwinds server I see it crash snmpd. The rest of the time snmpd runs fine.

It's only 4 of 12 servers which have the issue, and the only snmp related difference between them is about 4 years ago we loaded some mibs for a Velocity software trial.

I enabled dubug logging for snmpd and reproduced the issue. I see the CPU spike up to about 70% during the poll and then it dies. The last entries in the debug log appear to be related to Velocity.

Nothing is captured in /var/log/messages for snmp when it dies. I contacted Velocity software regarding the issue and they informed me that it's unlikely their mibs are the cause since Suse has been including their mibs by default in their net-snmp package since the GA release of 11.

I can post contents of the dubug log if requested, any suggestions?

Load default MIB, try to reproduce error. If you can't then you know it's likely the added Velocity OIDs in your customized MIB. (BTW "software trial" combined with "4 years ago" kind of leaves the impression of using something and not paying for it so maybe you want to clarify that ;-p) BTW[1], saying "Nothing is captured in /var/log/messages for snmp when it dies." kind of doesn't mean much as logging depends on where and what you log (see 'man snmpcmd', the "-L.*$" options). If it even won't log clues at debug priority then there's nothing else but to (verbosely) strace the process IMO.

Thanks for the reply. Finally after more digging I found this command which lead me to find an additional snmpd.conf defined in /usr/share/snmp/.

In this snmpd.conf I found old Velocity and Omegamon OID's. These were used to test out some monitoring software long ago and the source of my issue. I simply renamed this file to have it ignored, restarted snmpd and tried to recreate the issue; no more high CPU or snmpd crashes.