Refurbished Lenovo Thinkcentre M900 Tiny, in a bhyve vm on FreeBSD
this post was submitted on 03 Apr 2026
19 points (95.2% liked)
homeassistant
19027 readers
122 users here now
Home Assistant is open source home automation that puts local control and privacy first.
Powered by a worldwide community of tinkerers and DIY enthusiasts.
Home Assistant can be self-installed on ProxMox, Raspberry Pi, or even purchased pre-installed: Home Assistant: Installation
Discussion of Home-Assistant adjacent topics is absolutely fine, within reason.
If you're not sure, DM @GreatAlbatross@feddit.uk
founded 2 years ago
MODERATORS
I got a second hand HP Elitedesk mini from ebay. they are small and quite affordable.
I run way to many stuff on it.
Old desktops that the it guy at a previous job showed me how to request
@ohlaph Home Assistant OS on a Dell Micro with an i5-6500T in it and 16 GB of RAM.
Runs extremely well, just slow for ESPHome builds so I don't use the add-on anymore. Also while TTS is plenty fast I couldn't use any larger than tiny-int8 or base-int8 for faster-whisper. I offloaded that to my server with my old RTX 2070 in it and have it able to run the turbo model for speech to text.
But no Ollama or similar, fuck using those. I've only ever gotten uselessness out of them and I ain't paying someone else to use theirs to do the same thing just with slightly fewer incidents of "I didn't find a device called <the thing you said but slightly out of order and now the exact same as it's actually called>".
Home Assistant is in a VM in proxmox on a dell micro i5 8th gen. Zeave adapter passed through.
An "old" PC with an i7-4790T and 32 GB RAM.
I have also some Odroid devices based on 32bit-ARM.
But 32bit-ARM has the problem, that meanwhile many container images doesn't support this architecture anymore.
So, when your Pi is already 64bit-ARM it could be ok.
Otherwise the possible selection regarding available prepared container images may be smaller.
HomeAssistant Yellow
I use a dedicated Raspberry Pi (5, previously had on a 4).
I host everything else on a different server, the HA one is dedicated. Pretty nice because then it can run HAOS and basically manages everything itself.
One factor in keeping it separate was I wanted it to be resilient. I don't want stuff to stop working if I restart my server or if the server dies for some reason. My messing around on my server is isolated from my smart home.
I also have a separate Pi (4, previously on a Pi 1B) that runs Pi-hole, on it's own Pi for the same reason - if it stops working or even pauses for a moment, the internet stops working.
Yeah I run HAOS in a VM but I keep a backup on an SD card that I can pop into a raspi if for whatever reason the server is down.
People throw a lot of shade at the Pi but I love having dedicated hardware for some more critical projects.
Pi4 w/ SSD.
I host on a raspberry pi 4 in a Docker container. Ive added an ssd to the pi for longevity!
Custom SFF PC. Ryzen 2700X, Gigabyte B450i, Intel A380, and some WD red plus drives.
A supermicro 5018A-FNT4 with 16GB RAM. HA takes up about 25%, the rest is influxdb and Grafana.
Up until a couple of weeks I was running it on a dedicated Pi4. It's now running as a VM in ProxMox on a pair of Lenovo M710q mini PCs I got off ebay for £40 each.
I did load them up with RAM, upgrade the CPUs and add a second NIC so they probably came in at more like the cost of a 16Gb Pi5. Each. The RAM was the pricey part. I've measured the power usage and they each use about a 3rd more power than the Pi did which I'm happy with. Given that, the added flexibility of running ProxMox and how quiet they are I'm super happy with the setup.
Oh and I used to run PiHole on another Pi. That's gone now replaced with Technitium DNS running as a pair of VMs too. That was surprisingly easy to do.
It's now running as a VM in ProxMox on a pair of Lenovo M710q mini PCs
So, have you got High Availability setup? If so, I'd like to know more about that part...
So my plan had been to set up a pair of ProxMox hosts, use Ceph to do the shared storage and use HA so VMs could magically move around if a host died. However, I discovered Ceph and HA need a minimum of 3 hosts. HA can be done if you set up a Pi or some other 3rd host that can act as the 3rd vote in the event of a failure but as I didn't have Ceph I've not bothered trying.
I've read Ceph can work on 2 but not well or reliably.
I might setup a 3rd host some day but it seems a bit of a waste as I just don't need that amount of resources for what I'm running.
And I should have known really, I've a bit of a background in VMware, albeit at the enterprise level so I've never had to even think about 2 or 3 node clusters.
Ah, ok.
Yep, I'm in a similar situation... I have a few VMs, but not enough for lots of failover infrastructure... (redundant switches, etc)
I was thinking you might be just cloning 1 device to the other or something.
You can do HA in Proxmox with ZFS replication instead of Ceph. Third device something else as you said. It's what I'm doing.
Thanks, I'll look into it.
Please note that Home Assistant is officially supported on Raspberry Pi 4 and 5 with 2GB of RAM minimum Raspberry Pi - Home Assistant
If your older rpi is for instance a rpi 3 with 512MB of RAM, I'm not sure it's going to cut it.
Huh, I'm running it on a rpi 3B which was (barely) supported when I installed Home Assistant on it. It has only 1GB of memory but it's still working very well! I don't have a ton of automations though
Can confirm: Using a rpi4 with 2GB for a long time worked well.
N100. Cheap enough (last year) and plenty of power to run things like jellyfin on it as well. linuxserver makes great docker images.
Same setup, all in Proxmox.
Works really well
A Xeon E5-2650v4 on a Supermicro X10DRL-i and like a million dollars worth (128GB) of DDR4.
I have a proxmox host on a HP Elitedesk G3 with an i5 7500. In that I have a VM with HAOS and it runs like a dream. If something goes horribly wrong I can get remote terminal access from the proxmox interface along with rebooting and backing things up.
Also, you can actually run Grafana under HomeAssistant directly, though that does mean if HA is down then you also lose Grafana at the same time. IMO it is reasonable to use lots of stuff alongside HA but monitoring and remote access should be on a separate machine, and for that I have an old laptop (integrated battery so no need for UPS) and that machine is really only for managing remote access and monitoring.
That makes sense. And definitely need to keep my grafana service as is since it's basically a fitness service that I pair with Garmin. I was thinking about spinning up another instance to keep them separate since I might move the fitness service to a cloud provider eventually. Thanks for sharing!!!
Oh wow, I just had a look at that, normally Grafana is used for monitoring things like server response times and internal stats. Using for a Garmin fitness device is awesome! I would never have thought of it as a good way to get that kind of data and see it visually.
Do you use it to see your training progress over time? Or is it more for seeing specific runs and comparing? How do you actually use it? Is it useful for you?
In Docker on a Synology DS1522+, works well. I used to use the Synology HA app, but it was always some versions behind, and I was pleasantly surprised that backup and restore was easy to move it to Docker. I'd say that if you change your mind about how you host it in future that it will be fairly easy to change.
NUC i7, 32gb ram. Full Docker stack includes home assistant and all relevant containers for it zigbee2mqtt, esphome, vscodium, sqliteweb, rtl_433, mqtt, kokoro_tts, weaviate, zwavejs, openwebui, ollama, paperless with ai capabilities, a1111, whisper, sync thing, searxng, redis, qdrant, postgres. Runs fine, however ollama starts to max it out as I want to go for bigger models, so looking for something with serious gpu oomphbut still small footprint and low power consumption
VM under TrueNAS
3rd or 4th gen Intel NUC with an i5 and 16GB RAM. Running proxmox with 1 VM (Home Assistant) and about 10 LXC.
raspi 4 in a container.
basically running since the 4 was released
That's what I was thinking about doing too.
I moved home assistant from a raspberry pi 4 to a NUC a few years back with a few other services (ZigBee, mqtt, etc) and it's been fantastic. I'm also running frigate on the same system with a Google Coral for object detection and it handles the load of several network camera feeds well.
It’s hosted on Docker on Alpine Linux on a Raspberry Pi 4. It’s actually doing a trillion other things too and it’s not bogged down at all.
Put my old PC in a 4U server case and have it in a rack (minus the GPU)... I also added a ton of RAM early last year (thank god). I'm probably wasting electricity but it's fine.. I can spin up anything and everything without issues which is nice compared to my old NUC and laptop servers that would crash frequently.
I'm hosting on an desktop PC. Basically my old machine, but freshly installed with Ubuntu server and a few HDDs added.
I upgraded from a Pi2 to a second hand thinkpad. It went from underpowered to also running a NAS and a factorio server with plenty of precessing power to spare. I've used it for several other projects as needed. Docker compose makes everything trivial.
A Dell 3140 laptop. With HA and a number of other apps it loafs along at about 5% CPU. When Frigate is enabled it jumps up to 14% or so while drawing between 6 and 12 watts. The battery prevents power blinks in our area from being a problem, unlike my Pi4 that crashed and didn't recover when I was hundreds of miles away. The battery lasts hours when the power fails and the keyboard and display are convenient.
For those concerned about the battery, it's firmware limited to a 70% charge so it will last years. Current Li-Ion batteries have about a 1 in a million chance of fire. A car parked in your garage is hundreds of times more of a risk.
It's been excellent and I'd buy it again.
I have a HP microserver running Debian and Docker. But it runs far more than just HA. It runs all my self hosted stuff except for my pihole.
Dedicated N100 mini pc. It has more than enough everything…. Except for whisper and other AI stuff. That’s a silly situation where whisper takes longer than the dedicated gpu (and mind you we’re talking RX 6400, not 5090) outputs the results.
Pi5 runs my reverse proxy and then all services run as docker containers on my battery stripped laptop I used in 2020.
The laptop mounts a raid5 HDD array that the pi5 maintains