[c-nsp] Leaking global into VRF

[Jason Lixfeld]

On 2010-11-09, at 1:18 AM, Oliver Boehmer (oboehmer) wrote:

> Jason,
> 
>> I'm trying to lab up a scenario where I can leak routes from the
> global
>> table into a VRF, but I'm running up against an issue and I'm hoping
> someone
>> here can point out where I might be misstepping.
>> 
>> My P router is also my peering router.  That is, in addition to it's P
>> duties, it also speaks eBGP to another autonomous system.  I want to
> take
>> the eBGP learned prefixes and import them into a VRF.  This part seems
> to
>> work, but the issue is that the adjacent PE doesn't seem to see the
> prefix
>> that has been imported.  The PE sees the global entry, but it doesn't
> see
>> the prefix in the vpnv4 AF for the VRF in question.
> 
> This looks expected as a PE router (your peering router) importing a
> prefix from another VRF (or from global in your case) into a VRF never
> exports this prefix from the importing VRF into vpnv4. So in your case,
> you need the "import ipv4 unicast map VRF-IMPORT" on all PE routers
> needing the prefix.  

Interesting.  I was of the belief that MPBGP would take care of announcing these prefixes once leaked into a VRF AF.  Have I misunderstood the extent of MPBGP here, or is there another way to do it that uses (MP)BGP in some way?

Until then, I've set import ipv4 ... on all the PEs down the line, and while the prefix is now seen inside the VRF on all the devices I expect it to, my packets still don't seem to be getting to where I want them to go.  That is, they seem to be going nowhere.  I think one reason why is because no routers inside my network have a label associated with the eBGP prefix I'm trying to reach:

P1#show ip route vrf INTERNET 7.7.7.7

Routing Table: INTERNET
Routing entry for 7.7.7.7/32
  Known via "bgp 6666", distance 20, metric 0
  Tag 1, type external
  Last update from 7.0.0.1 00:02:38 ago
  Routing Descriptor Blocks:
  * 7.0.0.1 (default), from 7.0.0.1, 00:02:38 ago
      Route metric is 0, traffic share count is 1
      AS Hops 1
      Route tag 1
      MPLS label: none
P1#

And if this is potentially the root cause, how to get a label on this prefix isn't clear to me.  This is an eBGP prefix from an outside AS.  They have no knowledge that their announcements are ultimately going to end up in a VRF once they get over to us.  I only mention that incase it turns out to be part of the problem.