Ran into a “bug” last week while upgrading vCNS to NSX v6.1.5. It turns out that if you had VXLAN already configured, the sequence that the package installs its’ components is out of order. That’s one part of the bug, the other shows that the hosts are ready within the NSX manager when it really isn’t, so it is reporting the status incorrectly. After raising a case with P1 support, it turns out is a known bug, and the only remedy for it with this version is to remove the host out of the cluster, reboot, add it back to the the cluster, and reboot again. This procedure allows any wrongly installed components to be completely removed and then installed correctly once adding back to the cluster.
As you can see, this is a real pain and very time consuming.