When configuring an NSX-T Overlay network you have to increase the default MTU size of 1500. It is critical that the MTU is configured across the whole platform (Network Adapters, Distributed Switches, Physical Switches, NSX-T Uplink profiles). A typical MTU size is 1600 or 9000. Be careful when using high MTU sizes, as most of the drivers used in the VMKUSB NIC Fling do not support MTU size 9000 and thus, the overlay communication will silently fail.
The relevant part is when you change the MTU Size in the dvSwitch configuration.
When you change the MTU size, the configuration is pushed to all connected network interfaces. Make sure to verify that the NIC has actually configured to the correct MTU:
[root@esx4:~] esxcfg-nics -l Name PCI Driver Link Speed Duplex MAC Address MTU Description vmnic0 0000:00:1f.6 ne1000 Up 1000Mbps Full 00:1f:c6:9c:47:13 1500 Intel Corporation Ethernet Connection (2) I219-LM vusb0 Pseudo uether Up 1000Mbps Full 00:24:9b:1a:bd:18 1600 ASIX Elec. Corp. AX88179 vusb1 Pseudo uether Up 1000Mbps Full 00:24:9b:1a:bd:19 1500 ASIX Elec. Corp. AX88179
In that case, the vusb0 adapter has been configured to 1600 which is fine and sufficient for NSX-T.
If you try to change the MTU size to 9000, you do not see any error messages on the dvSwitch, but the vmkernel.log reveals that the MTU could not be set:
2020-07-19T16:10:42.344Z cpu6:524356)WARNING: vmkusb: Set MTU 9000 is not supported: Failure
2020-07-19T16:10:42.344Z cpu6:524356)WARNING: Uplink: 16632: Failed to set MTU to 9000 on vusb0