Here’s a great “gotcha” we ran into, maybe google will save somebody else this headache:
Two nexus 5k’s with a vpc-peer link between them. Each of them was getting a 2k connected with two fex ports. The first 2k comes up with no problems. The second 2k says “link not connected.” Of course we assume that the 2k either has an sfp instead of fex or there’s a physical cable problem. Nope.
Both 5k’s were using the same port-channel number (in this case 101). Going into the second 5k and removing port-channel 101 and building port-channel 1101 (keeping the same fex info) causes the interfaces to immediately come up.
Cisco, I understand that there are going to be weird things that come up with new architectures, but this works on 7k’s. Also, “link not connected” is totally the wrong interface status to show here. Maybe an err-disabled or something would be much more appropriate.