BGP and loopback interface under migration

[Marius Gunnerud]

Hi.

We are in the final stages before migrating to a new network.  For a short period of time (maybe a week) we will be running the two networks parallel to eachother.  All layer 3 interfaces...execpt for Loopback 100... on the old network will be shut down.  Loopback 100 is only used for BGP neighbor establishment in both networks.  The issue is that we were asked to keep the existing configuration in the new network, so loopback 100 has the same IP in both the old and new networks.

there will be connected a layer 2 trunk between the old and new network.  I am uncertain if there will be an IP address conflict in this case.  Will the new network be aware of the old loopback 100 interface?

I am hoping to not have to reconfigure the BGP router-id and neighbor statements.

any insight is welcome.           

[Jon Marshall · Accepted solution]

Marius

If all L3 interfaces on the old  network are shut then there will  be nothing to advertise the loopback to the new network or so i would have thought. But i'm unclear so to why you are bothering to keep the loopback interface on the old network. Is it simply because although you have shut all L3  interfaces you don't actually want to power down the routers themselves.

If so then defintely yes, there should  be no conflict because there would be no way of the new network learning that IP address. There would have to be a L3 peering between the old and new networks and then the loopback would need to be advertised to the new network.

So if i understand correctly you have interconnected the old and new with a L2 trunk.  I'm assuming this trunk is being used for end devices to connect to the new network ie. you have switches still in the old network with devices attached and these will need access to the new network.

If i have misunderstood and i may have as i'm uncertain why you are shutting down all L3 interfaces or there is more to the topology please add further details.

Edit - i have made the asumption that all L3 interfaces in the new network will be using their old IPs from the old network. So if the BGP peering is EBGP and you are using the loopback for EBGP peering then the other end of the EBGP connection must have a route to the loopback via the real interface. Obviously if the new real interface changed IP that would break. More likely the loopback is used for IBGP peering. In this case you would need to advertise the loopback into your network (or use static routes) and so if you had to bring up some L3 interfaces in the old network there is a possibility of a new BGP router forming an IBGP peering with an old BGP router (new and old in terms of your network). However for this to happen you would still need to enable the L3 interfaces in the old network and run the networks simultaneoulsy so it is not a concern as long as you leave them in the shut state.

Jon

[Marius Gunnerud]

Thanks for you reply.

We only have iBGP neighbors in the network.  We have 3 sites and we will be migrating each site over 3 weekends.  We just want to check that the logical structure works before patching over all the servers as it will be a massive job migrating the servers and then having to rollback (if it goes that way).

I tested this in GNS3 and didn't have any issues, so I am quite sure this will work fine (even though GNS3 is a virtualized environment).

thanks again.