[c-nsp] CEF fun in SXF

Gert Doering gert at greenie.muc.de
Fri Dec 9 09:41:36 EST 2005


Hi,

On Fri, Dec 09, 2005 at 04:37:52PM +0200, Saku Ytti wrote:
> On (2005-12-08 22:24 +0100), Gert Doering wrote:
>  
> > *All* directly connected (!) customers had massive packet loss - about 
> > 50-80% - while packets to non-"connected" targets had no problems at all.
> 
>  When you see something like this in PFC boxes odds are FIB is not
> syncronized between PFC and MSFC. When there is no route in PFC,
> PFC punts packet to MSFC ratelimited to ~300pps or so, so that MSFC
> can generate proper ICMP message. However, if MSFC happens to have
> proper route, traffic is switched correctly by it (all the 300pps
> that have succeededly being punted, instead dropped in hardware)

This scenario matches the observed behaviour very well.

>  What I'd suggest that you do is take 'show mls cef' and 'sh ip cef',
> take just the prefix part of the output, sort, uniq -c and print
> all lines that were found only once, these routes are missing in PFC
> and should have above problem.

Well, now it's gone, so right now we don't see any problems :) - but 
indeed, if it happens again, this is what I will try.

>  We've seen this issue in several 7600's IOS', but for us it's gone
> since SXE3.

Do you have any bug IDs or case numbers I could point 'em to?

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de


More information about the cisco-nsp mailing list