1. Had three hosts in one cluster, two of them with GPU cards NVIDIA GRID K1 and K2 and one normal host.
2. Deployed a VM with K2 vgpu.
3. Stopped the VM and tried starting it on a different host. (Expected it to fail as there is only one host supporting that particular vgpu type)
4. The log showed it as failed operation, VM remained in the stopped state, yet the notification showed "Start Instance" as successful and no error message was displayed.