We are trying to engineer the following scenario:
We have 2 HQ sites. The first is a primary data center and the second one is a DR site.
We have parallel MPLS networks from 2 separate providers.
We have points of presence into both MPLS networks at both of the HQ sites. We have separate EIGRP A/S's at the HQ sites and a transit EIGRP A/S between them. Our backdoor link is a Gigabit SONET connection.
At our remote cities we have one CE homed to MPLS provider A at location 1 and we have another CE homed to MPLS provider B at location 2. The remotes cities share a common EIGRP A/S specific to that city. The backdoor links are typrically Metro E using the same EIGRP A/S for transit.
What we are hoping to accomplish is we want to use both MPLS networks. We would like the routes local to remote location 1 to prefer MPLS provider A and remote location 2 to prefer MPLS provider B while at the same time having each remote site serve as a failover for the other.
On the HQ side our plan is to use HQ Site 1 as the primary point of presence for MPLS network A and use HQ Site 2 as the primary point of presence for MPLS network B. While at the same time having failover capability for each MPLS provider at each HQ site.
Of course we want the return traffic to the remote cities to use the same path they ingressed (don't want circular traffic flows from MPLS A via HQ to MPLS B, etc)
We are working from a clean sheet of paper. Our basic configurations right now just do basic route redistribution between BGP and EIGRP and vice verse.