[c-nsp] MPLS labels with VPNv4 blackholing

[Ross Halliday]

Hi Oli,

> -----Original Message-----
> From: Oliver Boehmer (oboehmer) [mailto:oboehmer at cisco.com]
> Sent: Tuesday, June 05, 2012 2:12 PM
> To: Ross Halliday; cisco-nsp at puck.nether.net
> Subject: RE: [c-nsp] MPLS labels with VPNv4 blackholing
> 
> well.. strictly speaking then you turn the RR into a PE as you have a
> VRF configured and inject prefix there.. I would possibly investigate
> using a distinct router, just to keep the functionality/features on the
> RR clean.. defining a VRF on a vpnv4 RR is quite unusual (not
> impossible, but very seldomly deployed).

In the strictest sense, yes, but if one were so inclined one could also state that any RR is a PE simply by the fact that it's running BGP :) Our network is very small - I can actually remember all the IPs of the core routers and we don't have any full P routers - so a distinct RTBH router is not a very realistic option for us. Currently we don't even have RRs, I'm actually building them right now

> Well, the FEC for vpnv4 is defined by the originator of the routes, and
> with per-prefix the FEC is defined to be different ;-)
> 
> So if you're worried about the distinct labels, you need to originate
> the pfx on a platform/version which supports per-vrf label feature.
> Looks like recent 15.1S images support per-vrf labels on the c7200..

Excellent - I hadn't thought of newer software. Been used to looking at the 6500 architecture for so long that I just assumed that if it was a platform-specific feature, nothing else would work the same way! With a typical 1:1 prefix:interface relationship I had gotten it into my head that labels were actually based on physical forwarding path, and that the per-VRF labeling on 6500s invoked some dark magic with pyres and chanting in the TCAM. ;)

Thanks for your help

Cheers
Ross