[c-nsp] BGP quandry

[Peter Rathlev]

On Wed, 2009-06-17 at 22:54 -0500, Justin Shore wrote:
> So I did a little more playing around and found that if I added an vpnv4 
> activate on the DC #2 router for core #2's IP I got my vpnv4 routes.  I 
> only got those connected to core #2 though.  I had to add another 
> activate for core #1.  I'm assuming that core #2 sent those BGP routes 
> that it learned via iBGP from core #1 to DC #2 because of the RR config. 
>   Since I'm eliminating the iBGP RR config I have to complete the full 
> mesh to get the full set of routes.  That makes sense.

Core #2 doesn't have "route-reflector-client" configured towards the new
router, so it only sends it's own prefixes and prefixes from any RR
clients of it's own. That seems to make sense to me too.

> One thing that doesn't make sense at this point is why the ibgp-peer 
> peer-group config in the vpnv4 address-family wasn't sufficient enough 
> to enable the learning of vpnv4 routes.  Do peer-groups and vpnv4 config 
> not mix?  Trying to add the command "neighbor aaa.bbb.ccc.ddd 
> send-community extendeded" to any of the routers involved (where 
> aaa.bbb.ccc.ddd is a configured member of a peer-group) results in the 
> error:
> 
> % Invalid command for a peer-group member
> 
> To me that implies that some sort of interaction exists between vpnv4 
> config and peer-group config.  Can anyone add any input to this?

AFAIK you always have to activate the specific peers in the VPNv4
configuration for VPNv4 functionality. I.e. :

 router bgp 64512
  neighbor PG peer-group
  neighbor PG remote-as 65412
  neghibor 10.0.0.1 activate
  !
  address-family vpnv4
   neighbor 10.0.0.1 activate
   exit-address-family
  ! 
  exit
 !

VPNv4 and IPv4 mixes fine, but the activation is seperated so you can
run some IPv4 only peers, some VPNv4 only peers and some mixed peers.

Regards,
Peter