[c-nsp] eBGP via loopback

[Aaron]

Did you setup ebgp multihop since you are doing peering to the loopbacks?
Curious on why you would want to use the loopback instead of the interface
for ebgp. Definitely not the recommended way unless you are trying to load
balance on multiple links.
Not 100% clear, but do you a static route on each end pointing to the remote
loopback?

Aaron

On Thu, Jun 12, 2008 at 7:53 PM, Steve Bertrand <steve at ibctech.ca> wrote:

> Hi everyone,
>
> I'm having an issue delivering packets in a test environment that I have.
>
> To make it as simple as possible, I'll describe what I have with all links
> disconnected except for the problematic one. My inquiries are not really
> regarding the packet loss, but more about BGP implementation itself.
>
>
> coe router --------- fibre --------- cpe router
> ebgp multihop (loopback) ---------- ebgp multihop (loopback)
> default-orig ---->----->----->----- default points to coe loopback
> /29 over loopback ----<-----<------ advertise /29 via loopback
>
> Unfortunately, I'm not aware of any tcpdump type software on a Cisco, so I
> don't know exactly where the packets are failing.
>
> The devices can communicate back and forth whether it be over the actual
> interface PtP IPs, or the loopback addresses themselves.
>
> If I direct a packet in from a separate interface on COE, I receive ping
> responses from all IPs, on all interfaces on that router. Attempting to ping
> any IP on the CPE, the traffic goes in, but I see no evidence that it is
> attempting to be sent out to the CPE destination.
>
> I'm fairly certain I'm missing something quite silly here, but before I
> spend any more time trying to test out my own personal educational theories,
> I thought I'd ask opinion on two thoughts I'm having. I'm hoping that the
> fine engineers here can steer me in the right direction.
>
> - using loopback with eBGP could possibly have a detrimental effect by
> trying to re-establish a BGP session to the remote loopback via a different
> multi-homed path that already has an established eBGP session.
>
> - in contrast to iBGP where there are usually multiple physical paths to a
> peer's loopback, eBGP is designed for PtP, so if the interface, link or
> remote interface goes down, you want BGP to notice that anyway and rebuild
> the routing table around the problem
>
> The loopback isn't the problem in the packets getting through as far as I
> can tell, but I just wanted an opinion whether I'm on the right track here
> with the utilization of loopbacks, or if I've been simply thinking about
> this too hard.
>
> Regards,
>
> Steve
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>