I am trying to decide best way to implement reliable and predictable resilience in a multisite OSPF/BGP network.
2 main sites A and B(linked by a 100Mbps LES) each with a separate BGP AS and individual link to BGP WAN (the B site is about to be connected)
many other sites around globe linked over a BGP network (each with own AS)
sites A&B run 10.A.x.x/16 and 10.B.x.x/16
Within site A OSPF is run
within Site B OSPF is run (currently same OSPF as A)
other sites may also use OSPF in LAN (but not running OSPF directly over WAN
All sites need resilience
A is to get resilience via B (over LES) and vice versa if the local wan fails
I dont know if I should continue to run OSPF between sites A and B via LES or run BGP across this link and keep OSPF separate between A and B
Also should i use dynamic OSPF <> BGP redistributions or use hardcoded network statement in the BGP?
can anyone offer any ideas or advice please?
potential issues are complexity, possibility of routing loops / asymettric routing, predictability/stability.
each site should use local WAN link preferentially, and fall back to other site wan link in event of a failure.
any pointers to relevant design doc might be useful too.