You need to do the ping from the one ESXi Host where you have one VM hosted towards the other ESXi Host where the other (destination) VM is hosted on.
You start with logging into SSH to both ESXi hosts and issue the command (like Richard__R proposed):
ping ++netstack=vxlan -d -s 1572 <dest VTEP IP>
If this ping DOES go trough (both sides) you at least know that your underlying infrastructure in terms of multicast (if you do not use the controller) / MTU is ok.
If the ping does NOT go trough you can try:
ping ++netstack=vxlan -d -s 1472 <dest VTEP IP>
Then you know something is up with the MTU not being 1600 end-to-end.
Other things that I would try is well is rebooting (or turn off and on again) both VM's that you are trying to ping to and from.
I have experienced problems before as well that VM's on the same Logical Switch and on different hosts could not ping eachother and my "solve" was rebooting the VM's.
If your VM's have two NIC's with two default gateways configured this can also be a problem.
You need to disable the NIC that you are not using in the OS by bringing it down, and also uncheck it as being "connected" in vCenter Server.
Do this just for testing purpose and then you can see where you go from there...
