Super quick blog post as I keep studying away for CCIE DC.
NPV. I won't bore you too much with going through NPV again and will assume you already understand the basic idea behind it: That the entire switch is treated as a single client by upstream switches just with multiple FCID's, and that the upstream switch has to support NPIV.
Still with me? OK
So now the question is, we have the NP Port, which is the "uplink" from the NPV switch to the NPIV Switch (the rest of the network) and the F Ports on the NPV switch which just go to the servers. So how does the NPV know what uplinks to take if there are multiple uplinks? Can it load balance the traffic?
Unfortunately I don't have a Nexus 5k So I can't verify the below with absolute certainty, but according to the configuration guide available here:
" When a server interface is brought up, the NP
uplink interface with the minimum load is selected from the available NP
uplinks in the same VSAN as the server interface."
In other words, configure your NP Uplinks first then configure your server ports and the traffic will be load balanced across the links.
Further to this, traffic-maps can be used like static pinning to force traffic out a particular interface.
The only question remains, what happens if you add a new NP Uplink further down the track? Unfortunately this is not that nice, basically if you add a new NP Uplink, only servers that login/logout of the fabric will be assigned to that NP Uplink, you can however choose to configure the NPV to perform a disruptive interruption of traffic to re-load-balance the traffic from all servers across all the NP Uplinks. I can think of just a couple of reasons however why you wouldn't want to do this ;).