I currently am working on building up my smarthome setup and I’m looking for a stable solution that’s not going to break my bank. I would like to get a Home Assistant server up and running, but I feel like I am suffering from decision paralysis.
Here’s what I’m struggling with. I’ve heard using a raspberry pi isn’t really a good long-term solution (unless you’re using an SSD – but even then, it’s not that powerful). I currently have one pi running with a sd card running homebridge (not ha), so whatever solution I come up with, I’d love to fold that into it. My initial thought was: find an old computer and run homeassistant on it. I have an old mac (13 inch, 2011), but because it has no battery, it’s CPU is severely throttled. Running my VM on it with HA probably wont last. For example, I have two cores dedicated to HA and even minimal tasks like running visual studio are taking 80% of the dedicated CPU. Maybe replacing the battery could solve this, but it’s also on an old HHD, so I’m guessing I’ll need to replace that. Heck, I also have to buy it off the guy I’m borrowing it from. I’m guessing that the total cost will be something like 250 dollars to restore it. Then, there’s some good reason to believe that it’ll cost maybe 30-40 a year to run all day and all night on power.
Alternatively, I could buy a NUC or something for the same price. I’m leaning that way, but one thing I’m concerned about is their ability to run multiple VMs? How feasible do you think it will be for them to run those?
What is a reasonable startup cost for a HA setup? Which machines would you recommend using. Would you use one computer and multiple vms or multiple computers each dedicated to a specific task (or, at least, just one focused on HA)?
I would definitely start looking at what your goals are.
If you are running home assistant and just home assistant I would really recommend what I am using which is just a rpi4 with an SSD. It is really reliable and you can do some addons as well. I am using doods which is a local object detection service and it uses around 5 percent of my CPU with the lite version (which is made aimed to be more efficient with some compromises on accuracy) and around 10% for the TensorFlow version which is extremely accurate. From my months of using it, it hasn’t given me any trouble. I got my pi with SSD for about $120 CAD
However, if you want a system that has cameras all recording videos, detecting motion and objects. then a beefier system is what you need. A nuc is a very good candidate because of its size and power.
If you are still wondering what to buy, this video may help.
Ok so basically if I’m just focused on basic smart home stuff (say a really customizable zigbee gateway, an ssd and a raspberry pi are plenty?)
On that note, for the pi, do I need more than 2gb of ram? And for ssd, does it matter if it’s exposed or not? Like I see that I can get one with usb built in, but those are more expensive than just getting an adapter with the exposed ssd. What about really a really cheap portable ssd. It seems like they’re all overkill on space and I’m trying to figure out if the price is for reliability or space. I saw one going for like 30 bucks, but is that just too good to be true?
Consider a previously-owned NUC to save some $$$. Remember (if you can find them) Pentium and Celeron NUCs will run HA absolutely perfectly for some time to come, with plenty of headroom for expansion later. Here’s one very good example:
Would this run ha as a virtualbox or would it struggle on that? That’s the concern I have because I want to run ha and other stuff like homebridge. And I really would rather not do docker because most tutorials and documentation I’m finding use supervisor, but maybe I just don’t need supervisor and don’t understand well enough.
I’ve been running my household and a couple remote rental units on a single pi3b+ for 4 years now without issue. It’s cheap, reliable, no moving parts and uses as much electricity as the nightlights in the hallway.
I’m thinking of upgrading to a pi4b soon… not for any reason other than “just 'cause”.
I run Home Assistant Supervised within Docker containers under Debian Buster on a similar i3 box. Docker works well, keeps things in separate environments, and isn’t difficult to learn, at all. Docker is actually a virtual environment. Each Docker container is it’s own entity, and they can communicate among themselves, when necessary. Home Assistant uses this quite a bit, and Supervisor makes installing add-ons an absolute BREEZE.
I’m using a 4GiB RPi4 + SSD, no performance issues. I have a handful of addons (mqtt, esphome, grocy, syncthing) and the pi is mostly idle. Unless you plan to run a ton of CPU heavy stuff, I think a NUC etc are overkill.
I am in a similar situation. Currently I got home assistant OS running on a Raspberry 4 with 2GB of Ram. My Plex Server is running on a Shield TV. Additionally I have a Synology DS416j NAS, which runs MariaDB and keeps the Home Assistant database to spare the SD card of the Pi with some more lifetime.
As I am gaining more experience with Home Assistant, I was looking into InfluxDB and Grafana. Unfortunately this cannot run on the simple NAS. It has got no Docker support either. Therefore looking to beef up my Home Assistant setup. A NUC could potentially also run the Plex Server, but looking at cost and energy consumption expanding the Raspberry Pi with an SSD might be the best way forward.
Now I have read about two ways to achieve this:
Run Home Assistant OS as per SD card on the SSD
Run Debian on the Raspberry with Home Assistant supervised in a docker container.
I am using an 8gb pi and it’s using about 25% of ram with Mqtt, motioneye, DOODS, advised etc. I have read somewhere that 4gb pi4 is plenty for home assistant + add-ons. For the SSD I recommend buying an SSD and a usb 3 to SATA adapter because there are some compatibility issues going around with pi4s and SSD adapters. You can see more about it here. Installing Home Assistant on a RPi 4b 8GB with SSD boot
A 128gb SSD is the most cheapest and reliable as far as I know
I installed Debian first, docker second, and then used the Hass.IO (as it was previously called) commands to install Home Assistant. Back then, installing Docker was the most difficult part, only because there were multiple versions of it available. Why Debian? Because I’m a UNIX/Linux command-line kind of guy. Why Docker? Because I had previously used it on the RPi3 and was comfortable with it. Why Hass.IO? Because it offered advantages (like add-ons, snapshots, etc.)
Ok that’s what I understood. So it sounds like the easiest way to run with supervisor is to use a dedicated machine or a VM. Are there things you can’t do without the add-ons that you can do with them? Obviously, it’s easier, but are there differences in what you’re allowed to do?
This is basically the only supported way.
Other install methods may lead to problems
Addons are easily configured from UI. Doesn’t require as much knowledge of service. Makes updates easier
All addons are available separately and may be configured and installed independent of HA.
Which method you use effectively comes down to preference
The best advice so far is to define your goals. The choice of running under a version of Linux and Docker vs installing a packaged solution that is offed for the PI or something like the Blue comes down to where you want to invest your time. I went for the Blue because everything is taken care of for me so my time investment is automating things vs spending time installing, upgrading he OS, components, etc.
Well actually only one. I mainly want to be able to run all my smart home stuff off of the one device (so mostly home assistant, but I also use node-red, homebridge, etc.). I think it’s more like one VM and a normal operating system. However, maybe the better solution is to just get a second pi running home assistant (operating system) and then have the original pi running everything else.
