Hello. We have a campus network with ~750 switches. Some of our switches seem to have really high CPU utilization. Usually, this is our larger stacks, (we have a few 9-stacks), but sometimes it affects medium to small stacks (we've seen it on single switches!).
Our current primary thought is that 802.1x is killing our CPU (we have to re-authenticate hourly - large stacks have lots of 802.1x sessions).... but, show commands don't support that. Also, our worst offender is only a 5 member stack - the CPU utilization is so bad, when SSHing, it feels like we're going over a satellite link, when we have ~55ms RTT. We haven't found a great correlation between any of the affected stacks.
Some hurried research shows IGMP snooping to be a culprit for many people... but we can't turn it off. TCAM utilization is nowhere NEAR 100%. And, according to the switch, the LED process is taking a huge chunk of CPU cycles.
So, what is YOUR CPU utilization? What's typical?
No comments:
Post a Comment