Hello Giovanni
It’s important to note that you are combining several different features into this topology. On the one hand, you have vPC, which creates a single logical VTEP, and on the other hand, you have VXLAN spine and leaf interactions.
When you create a vPC between two Nexus devices, you are creating a single logical device. This means that the two switches are viewed by other devices on the topology as a single switch with a single IP address.
Whatever features you enable on those two devices will function as if they are a single device, including the VTEP mechanisms.
Now let’s get to your topology. It looks like something is not quite right with the devices you have in each “VTEP” square. By definition, a vPC can be established between only two switches. Here, it seems like you have four devices in your VTEP box. I’m assuming the two devices on the bottom of each square should actually be outside of the VTEP box.
Secondly, it looks like you have some sort of LAG between the VTEP1 box and the VTEP2 box, something that doesn’t adhere to the rules of vPC, or the rules of VXLAN, since VTEPs should not be directly connected to each other, as you have also stated.
Finally, you also have a logical vPC between a server at the bottom and the two VTEP boxes. Such links should only exist between the pair of devices acting as vPC peers, and this is not the case here.
Can you share with us the source of this particular diagram?
Ultimately, it is possible to use vPC with VXLAN, and it is a good idea. But some caution should be employed. Take a look at this excellent Cisco Live presentation that shows many details about vPC and how to deploy it in a VXLAN environment.
I hope this has been helpful!
Laz