[c-nsp] Cisco BFD and NSF

[Pekka Savola]

On Tue, 1 Aug 2006, Saku Ytti wrote:
> However, the reasoning the Katz/Ward seem to offer, appears to
> to be because they consider it's hard to guarantee BFD
> enough processing time in control-plane, but rather they seem to
> believe, that some kind of failures might go undetected, if BFD would be
> ran in same place where routing protocols are ran. Would be interesting
> to know, what these failures might be and under what type of situation
> we could cover more failure scenarios by moving BFD further away from
> routing-protocols.

If Graceful Restart is enabled, running BFD on forwarding plane can 
gain a certain degree of assurance of whether initiating graceful 
restart would result in a black hole or not.  E.g., if you peer has a 
power loss or crash, liveness messaging or lack thereof on control 
plane might still indicate that GR is OK, while on forwarding plane we 
get an explicit statement that GR is NOT OK as reachability to the 
forwarding plane was lost.

This ambiguity is one reason (AFAICS) why Graceful Restart isn't used 
all that much.

As you note, putting BFD solely on the forwarding plane (without any 
keepalive on the control plane) could have potential for some nasty 
cornercase issues when control and forwarding planes are out of sync 
though..

-- 
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings