So NX-OS 7.2 is supposed to have a fix for the issue which won't let you run a routing protocol over a vPC enabled Vlan, if that routing protocol uses multicast for neighbor keepalives. To those of that have keep dancing around this issue it's nice to know that Cisco has at least recognized that the problem is real, but what about if you are using a 9K?
The solution seems rather simple to me (but I've been called simple). Just put the neighbor command in NXOS for EIGRP and OSPF. When you specify a neighbor with the neighbor command, the box uses unicast for all keepalives, not multicast. That should fix the stability problem and let you run EIGRP or OSPF, over a vPC enabled VLan.
The fact is sometimes it's just not possible to run two sets of links between switches (one for vPC peering, and one for non-vPC or 802.1q peering), like if they customer has a Metro-E circuit, and they don't want to pay for a second (or third and fourth since they should have two for redundancy), just for Layer-3 routing to a site they already have (two) Gig or 10Gig connections too, and which transport Layer-3 unicast just fine.
I love Nexus, but this is an annoying omission. Cisco, Please give us back our neighbor command in NXOS.
