[c-nsp] Full net table too large for Sup720 already?

Euan Galloway euang+cisco-nsp at lists.eusahues.co.uk
Sat Oct 27 09:48:07 EDT 2007


On Sat, Oct 27, 2007 at 02:32:26AM -0400, Afsheen Bigdeli wrote:
>  From this post:
> 
> http://puck.nether.net/pipermail/cisco-nsp/2006-August/032846.html

Don't believe that idiot ;-)

> It appears that the magic number, assuming you've tweaked the TCAM 
> appropriately, is somewhere between 244736 to 245546 routes.
> 
> I'd be interested to see what (if anything) happens when that number is 
> reached.

Well plenty of people have reached the untuned magic number already.

Depending on software version/features used/network design, you'll either 
get software switching of, or unreachability to prefixes that don't 
fit into the TCAM. Amusingly you software switch / lose "big" prefixes first.
So networks that just announce a /8 disappear, but those announcing every /24.. FINE.

http://puck.nether.net/pipermail/cisco-nsp/2007-June/041598.html

Shows the impact when full BGP table met default 192k tcam limit.
Looks like anything /8 - /14 would have gone a bit wrong unless covered 
by a more specific.

Oh the irony.

Double irony was that it was the fact that there *was* a default route that 
triggered the bug (if I'm remembering correctly, which I might not).

-- 
Euan Galloway


More information about the cisco-nsp mailing list