Yesterday I was working at a customer site installing EMC. After getting the switches setup and into EMC with SNMPV3 several of the switches 'went down'. They were still accessible, so looked and SNMP process was high. Also noticed ALOT of ELRP loop messages coming into logs from the uplink ports. Opened a case with GTAC, narrowed down the issue with snmp (elrp was causing the spiked cpu) they tracked down the loop, and snmp process goes down on the affected switches. My question was why were these non-looped switches affected by the loop, the reason was 'because they are all connected to the core'. That just doesn't seem to make sense to me. Loop PROTECTION should be keeping other switches from being affected, however in this case that doesn't happen. On the switch where the loop was at, ELRP shut down the ports as configured, just would broadcast out 'LOOP' and send messages to other switches.
So the main question is, is spiking the snmp process on non-looped switches REALLY by design? If so, why? I couldn't get an satisfactory response yesterday, so posing the question here.