[AusNOG] Cisco 7201 CPU hog

[Sam Silvester]

On Fri, Aug 10, 2012 at 9:27 AM, James Mcintosh
<james.mcintosh at rocketmail.com> wrote:
> Hi Noggers,
>
> We run a Cisco 7201 on our edge.
>
> It's currently running at over 70% CPU when doing 200Mbps. This seems
> awfully high to me for such a low level of throughput.
>
> Can anyone else here running 7201 give me some ideas or am I expecting too
> much from my 7201?
>

That does seem high. Depending on the workload, we can push up to
about 800Mbps. I know that Netflow and ACLs do tend to hurt these
boxes quite a bit though.

One thing to check is the output from the first line of "show proc
cpu" e.g: "CPU utilization for five seconds: 20%/19%; one minute: 20%;
five minutes: 20%"

The value for the 5 second counter after the slash (19% in the above
example) is the time the router is spending in interrupt (i.e.
forwarding traffic via CEF as I understand it).

If that value is significantly lower than the first number (the total
5 second CPU), then 'something' in the router is chewing up CPU at a
process level (non CEF traffic or the like - routing updates, that
kinda stuff). This link is a good start for debugging:
http://www.cisco.com/en/US/products/hw/routers/ps133/products_tech_note09186a00800a70f2.shtml

If it looks like the above with similar numbers, it's probably just a
function of the features you've got switched on. Check for what you
can minimise ACL wise. I've been led to believe the 'log' keyword in
ACLs may be a bad thing - I'm not sure though as we don't use that
keyword. Also see
http://www.cisco.com/en/US/products/hw/routers/ps359/products_tech_note09186a00801c2af0.shtml

I don't think the IOS makes a dramatic difference, but we've been
running 12.2(33)SRE train releases generally.

Cheers,

Sam