I've recently deployed a pair of Nexus 3064s in a VPC for core routing/switching. Since I've deployed these switches, we've experienced intermittent packet loss on our ESX devices. I suspect some kind of oddity on the ESX servers but I'm curious to know what is the correct configuration for the ESX hosts from a Load Balancing perspective?
Should our server team have them configured in Redundancy mode a.k.a Active/Passive links or should they configure as port-channel on their servers?
To drawn this picture out, the ESX servers are attached to a single 4948 within a rack but have multiple links on this switch. The 4948 is multihomed to the 3064s via a single port-channel with VPC enabled.
Redundancy mode vs port-channel is really dependent upon what the hosts support but in general port-channels provide you with additional bandwidth. I believe with ESX you may need to use mode on.
In regards to the intermittent packet loss, I would start with the basic: sho int count, check spanning tree, cpu, log, qos etc. Might have to do sniffer traces. Intermittent issues are the toughest ones to troubleshoot.