[c-nsp] disabling dCEF per VIP on 7500?

[Raymond, Steven]

> -----Original Message-----
> From: Gert Doering [mailto:gert at greenie.muc.de]
> Sent: Thursday, August 12, 2004 12:38 PM
> To: cisco-nsp at puck.nether.net
> Subject: [c-nsp] disabling dCEF per VIP on 7500?
> ... and I'm fairly sure that in a couple of months, their memory will
> be full (this is with all interfaces on "no ip route-cache 
> dist" already,
> didn't make a difference in memory usage).
> 
> Documentation says something like "in that case, dCEF will just be 
> disabled on the offending line card", but from experience, I 
> assume that
> it will result in funny things (CBUS complex, VIP2 restarts, or some
> other nasty service interrupting surprise).  So I want to turn off 
> dCEF for that whole line card *now*.

A few years ago, some of our routers had VIP2-40s w/ only 32MB.  When they
ran out of DRAM (as soon as BGP got about halfway loaded), you'd get a
malloc failure in the syslog, but everything stayed up and it was relatively
benign.  These boxes had a mix of 40s and 50s.  dCEF would stay working on
the 50s and "sh cef linecard" displayed that the 40s were "disabled".
Am not positive, but when dCEF disables, router then defaults to what, fast
switching?  Some non-distributed CEF?  The load on the RSP was definitely
higher.  In fact, when all the 40s were replaced with 50s, there was a nice
drop in RSP CPU.