We had swap cable in our network during the migration, BGP peering with ISP's loopback address over these swapped cable. My leader mentioned those swapped cable cause network outrage but from the lab, I found the traffic still be able to pass through even the wan ip is not in the same subnet. Please help!
Please see my lab test as below and let me know if there is anything wrong.
My topology has been attached.
RA and RB has ebgp seesion and peer with Loopback 0. Wan ips between RA and RB are in different subnet, I have configured static route on both RA and RB to point to each other's Loopback 0.
ip route 220.127.116.11 255.255.255.255 10.1.1.2
ip route 18.104.22.168 255.255.255.255 192.168.1.2
At this point, both routers has route to neighbor's loopback and I am able to ping RA's loopback WHEN sourcing from RB's loopback and EBGP came up.
RB and RC are IBGP peer with WAN ip, I have configured next-hop self on RB. The routes on RC learned from RB has RB wan ip as next hop. I have network command on RC for RC's loopback.
At this point RA has learned RC's loopback and if I announce RA's loopback address to BGP then RC also learn RA's loopback. I can get susseccfully ping from RC's loopback to RA's loopback.
Interesting result. First question I have is what link type you have between RA and RB.
If this were an ethernet link, it would probably not work because ARP will be unable to determine the mac address of the peer interface. However, if this is a serial line, with frame-relay for example, I do not see a reason why it should be impossible. As RB has a /30 on its wan port, it will send all traffic out that interface (pvc actually) without bothering about the ip address of the other side. The same holds true in opposite direction.
eBGP does then also work because you are using the loopback addresses as peer-ip.
This implies you already have ebgp multihop configured on both sides which is a requirement when not running eBGP between directly connected ip addresses.