[c-nsp] BGP/ISIS/administrative distance conundrum.

[Jason Lixfeld]

I've been trying to sort out a routing paradox in my lab and I'm hoping someone might have some insight.

Routers 1-4 are all speaking ISIS-L2.
Routers 1,2 originate a default route into ISIS.
Routers 1,2 speak MP-iBGP northbound and receive a default route from their respective northbound MP-iBGP sessions.
Routers 1,2 redistribute ISIS-L2 routes into MP-iBGP so the north side has reachability.

Problem - Because the AD of ISIS is lower than iBGP, R1 and R2 see the ISIS default route an ignore the BGP default route.  This causes a routing loop.

>From what I've read, it's not possible to apply some sort of filter to the the ISIS process on R1 or R2 to ignore a learned ISIS default route and just originate a default route.

I can't see any way of adding a route-map to the MP-iBGP sessions on R1 or R2 or either of their northbound MP-iBGP routers to change the AD of the default to less than 115.

I can modify the ISIS distance on R1 and R2 to be greater than iBGP, but that would ultimately lead to all ISIS routes on R1/R2 being ignored in favour of the MP-BGP routes for the same destinations being learned from northbound iBGP.  Likely more routing loops would ensue here.

I could somehow tag the ISIS routes that are redistributed into MP-BGP on R1/R2 and write a route-map to filter those routes from being announced back down to R1 and R2 alleviating the loop.  I'm wondering though, is there's a more straightforward answer that I'm not seeing.

Thanks in advance.