[nsp] ATM problems (DSL)

[Chris Roberts]

> >
> > On Fri, 4 Apr 2003, Chris Roberts wrote:
> >
> > >
> > > What routing protocols is this router running (if any?). Look for CPU
> spikes
> > > on the router with 'sh proc cpu' run a good few times over a five
minute
> > > period.
> >
> > BGP and OSPF.  Two BGP peers, total of maybe 20Mb/s.  I've got cricket
> > looking at cpu and it's not really spiking.  Both of our upstreams are
> > stable, with no bgp resets.  Our ospf area encompasses only 3 routers,
and
> > doesn't really change.

What type of area is it, normal or NSSA?

Use a tool such as 'mtr' (matt's traceroute) and use 'sh ip ospf stat' and
see if there's any correlation between the timing of the packet loss and the
timing of OSPF SPF calculations to see whether there is a large amount of
OSPF instability.

Using cricket or another tool to examine your CPU usage isn't very useful,
try 'sh proc cpu' (and run it regularly, say, every 10 seconds). CPU usage
can burst for a second or two when a routing update is received and then
drop back to normal, and these kind of spikes would easily get averaged out
with cricket unless you're sampling every second. It's these CPU spikes that
can cause the packet loss (or are a symptom of it).

Cheers,
Chris.