I am implementing a new edge architecture and in my lab everything seems to work right with the exception of the iBGP connection. A quick diagram is below:
CE-A -------- CE-B
CE-A and CE-B have an iBGP connection to propogate routes to each other from ISP-A and ISP-B from which I am receiving full routing tables. I am running OSPF between the CE's and Switches. The switches have a routed link between each other and trunks down to the LAN which is how they are propogating their HSRP multicasts for active/standby status. Each switch has two default routes, the first pointing to the next hop interface on its respective CE device and the second being a floating static in the instance that the link goes down to go towards its adjacent switch.
Both CE devices have routes pointing towards the VIP of each vlan on the switches. If the link goes down from lets say CE-A to SW-A CE-A reroutes to CE-B and down towards the VIP via OSPF. I have seen one flaw so far with this and that is that the iBGP session does not drop when the connection drops between CE-A and CE-B which causes some traffic to get blackholed. I attempted to use the neighbor x.x.x.x fall-over command but this didn't seem to speed up convergence.
Should I not be running OSPF at the CE devices and let iBGP do the work for rerouting traffic due to the static routes being installed in the routing table? Any other thoughts on where my thought process may be messed up on this one?