Whenever a rediscovery is run on a Nortel/Avaya 8600 ethernet routing switch, the CPU spikes to 95% for about 1-2 minutes. This did not happen prior to NPM 10.2
Hi HBrandNSB,
AFAIK these devices have lots of interfaces, so most probably this is caused by Topology data collection that happens during the rediscovery.
You can try to List resources for some device, remove L2 / L3 topology pollers from them and see if situation improves
thanks
I removed the L2/L3 topology from one of the devices and did a rediscovery and got the same results.
Hmm, then there is not much I can suggest here without deeper investigation. Please feel free to open support ticket in case this behavior cause problems for you.
Hi,
We're having the same problem with Cisco Catalyst 3750X-stack. Prior to 10.2 the CPU baseline was at about 17% and after the 10.2 upgrade the baseline jumped up to 57% and started spiking at rediscovery intervals. Changed the default value and the CPU graph changes accordingly. I have a ticket opened about it already. I was blaming Cisco on this because of the other problems with the SNMP but if Nortel has the same problems, it might be a problem with the NPM pollers after all.
Like I said, the 3750X stack seems to have also other difficulties in handling SNMP. The Toolset's switch port mapper won't work with it, because the IpNettoMedia (1.3.6.1.2.1.4.22) MIB is responding too slowly and the SPM will timeout before it's finished (maxed out the timeout). I can use SNMPWalk or MIB browser to get the data, but it causes the CPU to peak @ 100% during the poll. Our static routing is in this stack, but the polling worked fine when it was in a Catalyst 4503 Supervisor 2+TS.
We're using IOS 12.2(58)SE2.
- Lertsa
Just to clarify, removing the L2/L3 topology from discovery did actually resolve the issue. I don't use that feature, so its not an issue for me.
