Snippet - SDM templates, they kill kittens…
Earlier today there was an issue raised on one of our new(ish) ME3400 switches that we have started to deploy to customer sites.
We started getting SNMP traps from it complaining that its CPU was maxing out, not something that we would expect to see from a switch, let alone a switch that was WELL within its operating limits.
After jumping on sure enough the switch was showing a pretty high utilization on the CPU with regular spikes up to the mid 90% range.
After some regular diagnostics by the second line guys it got passed over and it was then that we saw the issue.
The ME3400 has two possible SDM templates, those being ‘Layer-2’ and ‘Default’ and it seems that this switch either came out of the box with ‘Layer-2’ enabled or someone enabled it during deployment (for some reason!?).
Usually having the wrong SDM template on a switch may just vary the amount of a particular amount of ‘things’ that you are allowed, for instance you may be allowed 2k route entries on a certain template but 8k on another etc.
With the ‘Layer-2’ template on the ME3400 however you get (amongst other things) NO space for IPv4 unicast routes which means that the TCAM has no space allocated for it, this is what was causing those horrible CPU spikes!
The ‘Default’ template that we later switched to has room enough for 8,000 route entries which is more than adequate!
For more information on the SDM templates on the ME3400 check out the Cisco page here (you will need a valid CCO login for it though).