Honestly? No. I think I just get really unlucky when it comes to buying batches of rPi4s.
THIS. People tend to forget that ‘add-ons’ are really just docker containers and each container requires resources above and beyond what running the add-on natively needs due to docker basically being a virtualization platform. It’s the same when people run VMs and forget that the host not only has to provide resources for the VM, but also the hypervisor that has to manage said host(s).
Continuation of the adventure: I rebuilt a brand new HA on the Latitude E6530 / SSD / 16 GB of RAM that I had in stock.
It’s day and night ! Nothing to do with Synology in runtime.
I didn’t make the mistake of putting all the containers back on this new server, but I was still happy to see the difference. It is incomparable. A restart will take less than 2 minutes, or even less than 1 hour!
I will therefore reassign my NAS for its primary function, storage!
Thank you for all the good advice that allowed me to take a step forward with Home Assistant.
HA used to use a very small power and memory and work great with RPI3. But recently I found that even on RPI 4 with my growing device and need it will not sustain HA. microSD will surely not recommended since you wont be able to have it running too long. Even running RPI4 + SSD Sata M2 is not very good when you have so many thing connected.
NAS + VM is not very great either, my NAS 918+ with 16GB memory and test it does not making it good. So recently I add a new ASRock 300X running Ryzen 4570G now that is something else all together. hahahahaha (I know its over kill)
Today I split my system into 3 server The Synology running several system like Plex, MQTT and my Ryzen running all the needed container for HA. My old RPI4 will be running a simple HA + all the require BT sensor/room assistant also acting as my proxy server.
However if you like to run RPI4 then go with M2 SSD instead of microSD and ditch the RPI3 (it wont last) hahahahahahaha
OTOH, running Hassio in an actual virtual machine, with limited RAM, and running a fair amount of addons inside Docker inside that virtual machine, as OP was doing, doesn’t sound like a very performant setup in any case.
I’m running Hass Core and about 13 other Docker containers (including 3 databases) on a relatively slow NAS and it’s not even breaking out a sweat.
That’s the way I am doing it to, and also the preferred way IMHO. If your HA VM gets borked, your other services will not suffer from downtime when you need to restore it.