[c-nsp] Mutual redistribution into and out of the MP-BGP superbackbone

[Jason Lixfeld]

If I redistribute OSPF routes into the MP-BGP/superbackbone from within a particular VRF, those OSPF routes, which are now BGP routes, get announced to any BGP speakers inside the same VRF.

Conversely, if I redistribute BGP routes from the MP-BGP/superbackbone into OSPF within a particular VRF, those BGP routes, which are now OSPF routes get advertised to all OSPF routes inside the same VRF.

Normally this would't be an issue, except in this particular case, the VRF where the mutual redistribution is happening happens to hold a full BGP table.

Two issues here.  1)  It's a bit silly to have an OSPF router see every prefix in the Internet routing table as an OSPF route.  2)  It's a bit silly for BGP sessions inside the same VRF to see every single OSPF prefix as a BGP route.

I can wrap the redistribute ospf x with a route-map and set local-AS, which will prevent eBGP speakers from receiving those OSPF routes.  iBGP speakers, however, will still receive those OSPF routes as BGP routes.  Setting no-advertise will likely break the superbackbone, so I don't imagine I can use that as a solution.

In the BGP into OSPF case, I'm drawing a blank on how to prevent the BGP routes from clogging up the RIB on other OSPF routers.  Again, any particular BCP or nerd knob?

I'm labbing this in IOS, but the redistribution, when moved to production, will happen on XR devices.

Thanks in advance for any ideas.