I have been running a variety of Intel NUC nodes in my vSphere homelab over the past 3 years now, including the D34010WYKH, DC3217IYE & DC53427HYE.
In that time I have unfortunately seen more than my fair share of USB drive failures and corruptions, generally with an error which looks something like this:
Error loading /k.b00
Fatal error: 33 (Inconsistent data)
These are not cheap and nasty, or freebie USB drives, so I would not normally expect to see this rate of failures. The error only occurs when you reboot the host, and the startup bombs out at the start of the hypervisor launch. I have often managed to recover the stick by copying back corrupted files from another instance, but generally I needed to rebuild and restore the image. An unnecessary pain in the rear!
The Root Cause
The NUC case can become quite warm during normal operation with or without the fans spinning up, and I have come to believe that the main reason for the corruptions is that the USB stick itself is getting too hot and therefore eventually failing. Having pulled a USB out from a recently shut down node, they are really quite hot to the touch. You don’t actually see the symptom / failure until a reboot because the ESXi image actually runs in memory, so is only loaded from the USB stick at boot time.
As for the solution, it’s really quite simple. I purchased a number of 12cm (5 inch) USB 2.0 extender cables on eBay for just 99p each (including delivery!).
These keep the USB stick indirectly attached to the NUC chassis, and as such the heat does not transfer into the flash drive. Since doing this I have not seen any further issues with the corruptions. Job done!