Or if you are working in bird now with default only, you can just filter out other routes from exporting to the kernel or other clients in export filters. On Thu, May 30, 2019 at 4:48 AM C. Jon Larsen <jlarsen@richweb.com> wrote:

Setup an EBGP multihop session off your route reflectors, assuming they are bird, and export all routes and import none on your session facing his router.

So your customer has an extra bgp session (or 2) for the full route feed. Your primary bgp session works as is, the multihop sessions provide the full route feed.

Works well. We use it so we dont have to carry full routes on the peering devices.

He should add static /32 routes for the ips of the route servers (again, could be your rr, or just any vm that has a full feed) pointing to his transit (i.e. to you).

On Thu, 30 May 2019, Kevin B wrote:

...

Hello,

My customer wants me to announce a full IPv4 table via BGP to him, but we don't use it on the BGP node where that customer is peered. We receive the default from our BGP upstream, but can actually ask for and receive a full view as well, but only for a purpose of forwarding it to the customer.

What would be a scheme to receive a full-table from a BGP upstream and forward it to a BGP downstream, without importing it to our own routing table?

I am not very experienced with Bird, but some guys on Freenode ##networking have suggested using a pipe, can someone help me on this please?