So maybe I got lucky here but thought I would share my story… I have been running HA on a simple intel NUC (NUC6CAYH) with Ubuntu / Docker since Feb 2019 and it has run well. At the time I thought I would run a few things on the NUC in Ubuntu alongside HA in Docker but that really never happened. The couple things I did run are now add-on in HA so I decided maybe it’s time to get rid of another layer that has to be managed, make it simple and move to a fully supported environment. I downloaded the NUC image and copied it to a 128G SSD with Etcher and swapped it out with the current SSD on the NUC. In less than an hour I was up and running with a config restore from a snapshot I did prior to the disk swap and copied to my Windows desktop. The first thing I noticed (maybe it’s just me) was how much more responsive the system was, everything seems visibly faster such as time from motion detection to time when the lights actually turn on. I have a Conbee Version 1 running native (No deConz) and a Aeotec Gen 5 both of which came right up. I am using some Dome Z-Wave and Smart-things Zigbee sensors along with integrations from Lutron, MyQ, TP-Link, Two Honeywell Thermostats, Yamaha Reciever, Octoprint, UPS NUT, Alexa and various Sylvania, Sengled and GLEDOPTO lights all of which came back with zero issues.
Again maybe I just got lucky but the move from Ubuntu / Docker to a native clean install on the NUC went seamless for me and I’m now running a skinny easy to maintain environment which appears to perform much better. My NUC just idles even for what it is it is overkill so if you do go to a NUC and all you are going to run is HA you don’t need much power.
4 Core Celeron w/8G Ram and 128G SSD (seems like a waste but highly stable)
Core CPU Usage 0.1% Core RAM Usage 2% Supervisor CPU Usage 0.7% Supervisor RAM Usage 4.8% Used Space 3.1%