I'm pondering the best way to utilize the redundant connections on the servers. I can't figure out what the best way would be. I think scrapping it (currently using RPVST and HSRP w/interface tracking) and going with EIGRP with the hosts behind the o, p, and q 3750s pointing to the 3750s as their gateways would work. Plus, I would have load-balanced connections between the switch blocks due to the equal cost paths. BUT!, what to do about the darn servers? If I use GLBP then traffic flow will sometimes be suboptimal. If a server wants to go to e but h is acting as the forwarder, it won't use the better path through g. Side Note: The servers have their interfaces bridged.
I realize that i, j, and f are not redundant. However, I'm really not too concerned, I added them to show the overall layout. What I am focused on is everything between f, i, and j. I would like the servers and the clients on switches o, p, and q, to have the most efficient redundancy.
The servers have not been connected yet. But, currently, g and h are running HSRP and tracking the non-STP-blocked interface to o, p, and q. g and h are alternately configured as primary and secondary for each of the vlans that o, p, and q are a part of (hope that makes sense!). Everything is running Rapid-PVST so the HSRP should be tracking the spanning-tree. If spanning-tree reconverges the HSRP routers will flip-flop.