Yes even compare to simple mini pc, the usually either work fine and fail completely. The Raspberry when it fail it isn’t necessarily obvious at first.
For the crashed hard drive it’s now less of an issue with the newest generation of ssd. And the scale is way smaller than with the sd card as you said.
I also believe the stability of the system depends on various factors such…
- How many occupants in the house? Are they always in the house or only back at night?
- How many sensors or components you install in your HA
- The number of automations and how frequent they are triggered.
- The complexity of each automations. I have automations that automate another automation that automate yet another automation.
Obviously, the system is more stable for someone who is living by himself in a small apartment compare to let say a big family with half a dozen of children and grandparents who stay in 10 bedrooms bungalow most of the time.
This might have an influence on how fast an instability would be uncovered. But if the system is stable, that you fire 5 or 200 event per day should not make a difference on the stability (but on performance and life length of hardware.).
I haven’t restarted home assistant since 1/2. Occasionally get a wink error (use it for my zigbee lights) right when I restart which is solved by another immediate restart. I also have an aeon z-wave stick with 15 devices, wifi bulbs, mqtt lights, and a few other components. I’ve found I only need to restart when I change some yaml.
Running raspbian on an rpi3 with a mysql database (I was having lots of issues with sqlite freezing up, solved by switching to mysql and only keeping 7 days of logs).
I used to have a problem after RPI restarts that were due to HA starting to come up before the network finished coming up. With the latest RPI updates, there is an option to have it wait till the network is up before it continued on. That helped with my DNS server and my HA server.
if biggest limitatyion of the raspberry is the SD card, what do you think of this system?
That’s great if you need portability. If you don’t, you can use the existing Pi3 with an external drive.
It must be as fine as any other mini pc with ssd storage.
So the pi3 with an external drive will work? I would rather use this as solution to my existing Pi3, do not want to rely on a SD card for too long
will the pi3 boot from a usb attached ssd drive
Been running 0.34.4 for 14 days, 23:05:03.419926 hours without issues.
https://www.raspberrypi.org/blog/pi-3-booting-part-i-usb-mass-storage-boot/
It’s a functionality that you’ll need to activate (only work on pi 3). Or you can simply use the sd card as a boot media (also work on pi 1 and 2):
1 Like
IMHO, you’re actually are better off booting from the SD Card and then just taking anything with a large volume of read/writes. Keep the SD card straight system and set up logs, DBs, etc. on the SSD/External USB/HDD.
Note that this opinion is based on theory and research only at this time; not on personal experience. While I’d like to do this myself it will either take a fair bit of research or a complete reinstall from scratch, neither of which I have the time for at the moment.
Isn’s a microSD to USB adaptor just a card reader like we used to use to get pictures off of our SD cards?
Yes, so it would be just trading one SD card for another. Probably add a little latency too.
But it could be useful for imaging your sd card, wouldn’t it?
I suppose so; I don’t know what you would use to create the image locally though and whether it would work while still mounted and in use.
I use the old Win32DiskImager method.
Old school huh 
. The problem with the ssd drives to run everything on, is how to get all the supporting files on the ssd instead of the SD card. Would you just have to setup links from the traditional directory locations to directories on the sd card like etc, var, usr, etc.
Maybe you can answer another question for me, that’s outside the scope of this question, but here it goes. The python virtual environments; they take all the python apps that are installed and save them in their own little virtual world. RIght? But the hass user has a home directory that isn’t in that world. RIght? Do non-pip installed apps go into that world if installed after the virtual environment has been activated? Do they go in there at all?
I was thinking of a microSD JACK to insert to the microSD slot, aqs adapter that convert it to a USB plug, where you insert an SSD.
But not sure if this exist or is possible technically …
That’s why I put that disclaimer out there. To do this I’d either have to put in a lot of research on how to move things and link things back or reinstall from scratch and STILL do a lot of research on making sure stuff was going to the right place.
I’ve only installed HA twice, once on my Win10 box and then moved over to a Pi. In both cases, I used the AIO installer - great because it simply worked and I didn’t have to baby sit the install.
Bad, because installing it in a more manual fashion probably would have helped me understand enough to validly answer your next questions. My understanding (which is admittedly weak in comparison to some of the other users/devs here) is that the hass user’s directory holds the user configurable objects and logs while all the programs and libraries, as well as the dependencies to run them, go in the VENV. So if you do an update, you know you have to activate the VENV so it can write everything to it.
Third Party stuff can be installed in the hass user’s directory. Most of the things I have added like the Harmony-API, AppDaemon, HADashboard, etc. are all installed in my pi user’s directory but run just fine. I’m sure someone will point out that I am doing this wrong, but it works fine and I haven’t had to set any permissions to a dangerous state to do so.