MPLS and next-hop-self

[lars.molstad]

Hi folks.

I came across a strange configuration of a MPLS network...

The network har two route-reflectors, and all PE's use the route-reflectors - so far it's OK, but... The route-reflectors has the next-hop-self set on all peers in the vpnv4 address-family... Wouldn't that force all traffic from the PE's to go through the Route-reflectors for rerouting?

L@rs

[mheusing]

Hi Lars,

In principle you would be right, IF the developers would not have been smart. If you configure it or not, a RR will not insert itself as next hop for any client. Makes a lot of sense (at least to the poor RRs ;-) In an MPLS environment, it really would break end-to-end LSPs and thus this single line of configuration could break your whole MPLS VPN BB. Once again, smart thing not to do it whatever the config says ;-)

Hope this helps!

Regards, Martin

[Harold Ritter]

Lars,

next-hop-self never affects the reflected updates on the RR. It only affects updates received via eBGP if the RR is also configured for eBGP obviously.

Configuring next-hop-self on the RR is useful in an IPv4 context when your RR is also configured for eBGP. It doesn't apply to VPNv4 though as next-hop-self is the default behavior for routes injected in VPNv4.

Hope this helps,