[c-nsp] Rationale for ISIS default origination behavior

[Saku Ytti]

On (2013-01-23 09:38 +0100), Adam Vitkovsky wrote:

> What would be the best trigger for a default route advertisement either from
> the edge(PE) routers or core(RR) routers. 
> -eBGP peer's interface might be reachable though the eBGP session might be
> down. 
> -ability or inability to see a single route (or couple of routes for that
> matter), might not reflect the ability of a particular edge to perform
> full-table routing. 

In this case you can have your cake and eat it.

ip route 0.0.0.0 0.0.0.0 8.8.8.8 
ip route 0.0.0.0 0.0.0.0 ASBR_ANYCAST 200

Or you could even have:

ip route 0.0.0.0 0.0.0.0 important_inet1
ip route 0.0.0.0 0.0.0.0 important_inet2 190
ip route 0.0.0.0 0.0.0.0 important_inetN 199
ip route 0.0.0.0 0.0.0.0 ASBR_ANYCAST 200


(ASBR_ANYCAST, is some (same) host/32 route in loopN in all of your ASBR, so
always resolves to closest ASBR up.)


So as long as important_inetN is advertised, you are using it, which
guarantees that edge is alive and is receiving at least this prefix.

If non of important_inetN is advertised, you can fallback to anycast ASBR
edge, hoping the edge works. 


But frankly, I'd opt for having just one important_inetN, some reasonable
stable destination. If that is gone, just fall back to closest ASBR
Anycast.

-- 
  ++ytti