[c-nsp] inet vrf

[Phil Mayers]

On 03/10/2010 08:30 PM, Tim Durack wrote:
> Having a hard time figuring this out. I have a vrf with around 30k
> prefixes (partial Internet.)
>
> Sup720-3BXL shows:
>
> RTR-2#sh mls cef summary detail
>
> Total routes:                                    58231
>      IPv4 unicast routes:                         29043
>          IPv4 non-vrf routes:                     89
>          IPv4 non-vrf routes (internal):          0
>          IPv4 vrf routes:                         28943
>          IPv4 vrf routes (internal):              11
>      IPv4 Multicast routes:                       5
>      MPLS routes:                                 28849
>      IPv6 unicast routes:                         330
>          IPv6 non-vrf routes:                     6
>          IPv6 non-vrf routes (internal):          0
>          IPv6 vrf routes:                         69
>          IPv6 vrf routes (internal):              255
>      IPv6 multicast routes:                       3
>      EoM routes:                                  1
>
> What are "IPv4 vrf routes" as opposed to "MPLS routes"?

The box will generate and advertise a distinct MPLS LSP for each route 
which has a next-hop, plus one overall label for the VRF aggregate, for 
"connected" routes.

That is, these 4 routes:

192.168.1.0/24 -> 192.168.100.100
192.168.2.0/24 -> 192.168.100.100
192.168.3.0/24 -> connected
192.168.4.0/24 -> connected

...will generate 3 MPLS labels:

192.168.1.0/24 -> label=98
192.168.2.0/24 -> label=99
192.168.3.0/24 -> VRF agg. label=100
192.168.4.0/24 -> VRF agg. label=100

So you'll burn a lot of MPLS labels

>
> Looks like the prefixes count twice from a fib perspective. Is that correct?
>