mpls/ldp question

If i have two directly connected bgp neighbors in mpls vpn l3 scenario (and they even peer with directly connected endpoints), why is it that -mpls ip- still needs to be enabled on that link, when there are no routers inbetween?

r1

Find more posts tagged with

Comments

marlon23

because your VPNv4 prefixes carry label with them to identify VPN, even you don't need outer label to get to the next-hop, you still need inner label to identify VPN. Hence you need to encapsulate to MPLS, so need MPLS running on the interface.

acidsatyr

marlon23 wrote: »

because your VPNv4 prefixes carry label with them to identify VPN, even you don't need outer label to get to the next-hop, you still need inner label to identify VPN. Hence you need to encapsulate to MPLS, so need MPLS running on the interface.

aha! got it.

Because i thought that that's what bgp does puts the inner VPN label. In fact if i do:
sh ip cef vrf VPN_A <some_cust_route> even without mpls enabled, cef still shows the label assigned by BGP.

So mpls is still needed to actualy "put" that inner label on?

Thanks so much.

edit: is there a way for bgp to do that, so that i don't run mpls but vpnv4 are still reachable between two directly connected bgp peers?