We are migrating from multiple locations as separate cores into a centralized core
Each location has many links and routers behind it, and each location runs its own OSPF area 0 network
Each of these locations has a main core router which terminates customer PPPoE sessions through VPLS to each edge router
These existing core routers will be turned into a distribution router, connecting to the new core and the PPPoE sessions will terminate on the new core
I am running iBGP between each main distribution router and the central core in order to keep the networks separate but still fully connected (we want failover so if i.e. the uplink from site A fails, it can route through a cross connect to site B then up to the core)
BGP keeps the networks separate, easily identifiable and I can manipulate the routing metrics easier. I don’t need Site C to know every individual route in Site A for instance. But Site C can still receive summarized routes from the core
The problem is we use LDP for VPLS tunnels, and LDP doesn’t work over BGP
I am not opposed to running OSPF on the core but not sure of the best way to go about it if that is the only solution. I would rather not
If I put the core into Area0 then it’ll join all the networks into 1 big area, slow down convergence and it’s a bit harder to control routing since every router must contain the full table. No summarization
I could use multiple instances of Area0 with redistribution where necessary (i’ve done this in area’s where it cross connects between sites) but I feel this can get kinda messy and not sure if it’s really an accepted practice
I could make each site a different area but then it becomes a fair bit of extra work with downtime reconfiguring all routers in each area
Is there a way to keep the existing LDP/VPLS infrastructure in place, but somehow keep BGP for the last leg from each distribution router to the core? (directly connected with a layer2 service)
We need VPLS tunnels from edge routers to the core run PPPoE sessions. So I can’t just use normal routing from each distribution site, it needs to be a continuous layer2 tunnel all the way to the core
My first thought is keep LDP on each distribution router and all its routers downstream, but then use BGP VPLS from the core to each distribution router, and then bridge the BGP VPLS tunnel and LDP VPLS tunnels. But then what happens if the upstream link from Site A fails? It won’t fail over to re-route through SiteB