[c-nsp] route-target import on non-leaking PEs

[Jason Lixfeld]

On 2012-12-13, at 9:56 AM, Phil Mayers <p.mayers at imperial.ac.uk> wrote:

> On 13/12/12 14:47, Jason Lixfeld wrote:
>>> 
>>> Yes. In fact, that's *required* if you want to do multi-path.
>> 
>> I seem to do multi-path just fine with maximum-paths ibgp 2 on my RR
>> clients inside a VRF that sees a default sourced from two different
>> RRs.  Said VRF has a common RD between the two PEs.
>> 
>> How is that different?
> 
> Well, AIUI multipath *ought* to require unique RDs. Obviously not; I wonder how that's working for you?

So based on the link you posted previously, which I am currently making my way through, what's happening is on my production network where this multi-path stuff is actually working, I'm using XR as my RRs, which has add-path support.  

> Basically, as I'm sure you are aware, without "add paths" support, BGP can only advertise one copy of a route - any "update" for a route implicitly "withdraw"s a previous route.
> 
> Therefore, if your RR sends:
> 
> update nh=x.x.x.x nlri=65000:1:192.0.2.0/24
> update nh=y.y.y.y nlri=65000:1:192.0.2.0/24
> 
> ...then the 2nd update should wipe the first. This is why (in theory) multipath requires a unique RD:
> 
> update nh=x.x.x.x nlri=65000:1:192.0.2.0/24
> update nh=y.y.y.y nlri=65000:2:192.0.2.0/24
> 
> ...are different routes.
> 
> Out of curiosity, what does:
> 
> sh ip bgp vpnv4 vrf XXX net/mask
> sh ip cef vrf XXX net/mask detail
> 
> ...say for one of your working multipath prefixes?
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/