[c-nsp] Cisco BFD and NSF

[Charles Spurgeon]

On Sun, Jul 30, 2006 at 10:23:31AM +0200, Mikael Abrahamsson wrote:
> On Sun, 30 Jul 2006, David J. Hughes wrote:
> 
> > BFD is such a great step forwards for high availability.  It's a shame 
> > that it appears to be a bit piecemeal at the moment.
> 
> We've seen BFD-flaps due to high CPU load on one end, so one should really 
> consider the risk/benefit ration when deploying it.

I second that experience. And I agree that BFD would be a great
feature if it worked reliably.

However, my tests on sup720s running sxf2 showed that BFD would
occasionally toggle the link when the "wri mem" command was run or
shut/no shut was done on an int or when the config register was
changed (both of which cause an automatic write to SP memory).

IIRC, this was with BFD timers ranging from 50 - 100 ms and packet
loss set at 3. This was an aggressive timing setup since I was trying
to get fast cutover across parallel links with dissimilar speeds
(1Gbps/10Gbps) in an IGP situation.

Apparently they are having problems with BFD access to the CPU during
various file writing events. Cisco is aware of the issue and I was
told that there were bugfixes in the pipeline. IIRC, the fixes were
committed to the 7600/12.2SR branch. 

-Charles

Charles E. Spurgeon / UTnet
UT Austin ITS / Networking
c.spurgeon at its.utexas.edu / 512.475.9265