[c-nsp] mpls on one interface carried over an l2tpv3 on another interface

[Joe Maimon]

Gert Doering wrote:
  >
> So if I understand this correctly, you need the L2TPv3 tunnel to build
> a "virtual ethernet" between two 7500s, that can then be used to tunnel
> MPLS across it, from two other interfaces on the same set of routers?

You understand correctly.

> 
> 
> 7500 fa0/0 -(spit out MPLS)-> same 7500, fa0/1 -(tunnel via L2TPv3)->
>   other 7500 -> ...?
> 
> 
> It *should* work, but I wouldn't be surprised if things like this confuse
> the ARP/CEF handling code of the box ("I can't learn that MAC address 
> *here*, because it's mine, and used *there*").
> 

Everything that works as you would expect on a 7500 surprises me.


> 
> This question doesn't really grok.  Where does VRF come into this?

As an example of simple vrf-mpls-vpn testing.

> 
> There's two things to diagnose this:
> 
>  - ping from 7500-1: fa0/0/0.2 <-> 7500-2: <mpls interface>
>    - there is *no* MPLS involved here, just plain IP, possibly inside
>      a VRF (depending on the 7500s are configured)

Works

>    - if that doesn't work, you have a problem with "bridging my own
>      interface over my other interface", or maybe with "subinterface".
> 
>  - if that *does* work, you should check whether LDP/TDP and your IGP come 
>    up on that link - which is then needed to transport MPLS frames.  But
>    for a link with MPLS frames on it, you normally would not put it into
>    a VRF, so this whole setup is confusing me somewhat.
> 

ldp/igp works (the tunnel is to carry the frames from the tag switched 
subint)

> Actually, if all you want is "transport VRF IP packets to router B", 
> this whole setup is much too complicated.  Just setup an IPIP or GRE
> tunnel, put it inside the VRF, and forward packets.

At that point I would need to redesign each vrf to skip the mpls 
cuteness. There is more than one and each has different routing policy.

> 
> gert

Thanks,

Joe