[On Hold] Deprecating Home Assistant Supervised on generic Linux

What is then Nodered, fileditor, duckdns, nginx, mariadb, zwave2mqtt,… ? People will have many many more problems with HA core and installing these separate containers and connecting them to HA, like problems now on supervised

It seems like the debate is around two types of users and I don’t know which is the majority and it probably doesn’t matter.
A number of us, early users, managing either our core installation on a VM or on a full machine with or without venv, manage our installations tightly and have no use of supervisor. A VM really is a lot simpler to manage for people who know how to, backing up the entire VM, taking snapshots of it are all very easy. Upgrading to exactly the version that you want with the component that you want through a single command line or worst case a shell script is easy enough. It has the perfect blend of having a full OS with full flexibility, is portable, does not interfere with the host OS environment.
The container is a bit extreme and I am actively trying to eliminate all of them. They add way too much complication and are too limiting for the benefit they offer but I understand that for some, the GUI, and mouse click action makes night and day difference. You just have to deal with the container’s overhead and the multiplication of individual environments. I just got rid of my unifi controller container and moved it to a VM. Upgrading was way too complicated with the docker container, forcing me every time to do a backup data, delete, download and create a new one, restore from backup, loss of uptime and data… Now within the VM, it is 2 command lines which I put into a script. It is 3s vs a 5min ordeal. It’s the same story with HA. So I am one of those who will never want to use the supervisor…
At the end of the day, the supervisor works for some and a large number of people wouldn’t run without it… But yeah, it is in no way essential. It is a complicated (but nice) solution to a simple problem.

I know there are ways to manually run a MQTT broker and connect it to HA (I had that setup on openHAB), and there are also ways of manually running a zigbee controller (as an alternative to the deCONZ plugin), and there would probably also be manual solutions for my other 7 plugins - but why should I “waste” so much of my rare spare time setting this up while the Supervisor can provide all of this with just a few mouse clicks, automatic updates included? And I really love the snapshot function! (I assume this is also not available in the Core only)

As I said before, for me, this is the main advantage compared to other competitors of HA and the reason why I switched from openHAB to HA a year ago. Things are just way easier and less time consuming, thanks to HA Supervisor.

I understand that maintaining the whole HA universe is a lot of work and I really appreciate this work (a big THANK YOU at this point to Paulus and the whole team!), but I would still find it really sad for HA if we would loose the supervised installation on generic Linux systems.

2 Likes

I would not know why they would have any problem… I actually find the separate containers much more problematic than running them all in one VM. I run Home Assistant, openLuup, Grafana, Z-way-server, Deconz, ha-bridge, sonos http server, nginz, unifi controller, many other servers and a bunch of HA custom components in the same VM. It’s much cleaner, easier to backup/maintain and resource efficient than containers… But I know it takes some basic linux CLI knowledge to do…

True, but

But with significantly increased complexity. The Supervisor implementation makes it extremely easy to add components such as Node-RED, that then integrate seamlessly, with virtually no knowledge of Docker or anything like that - it just works.

It might not be the end of the World, but it may be the end of HA for many.

Personally, I’d drop back to Node-RED. I don’t do anything super clever with the integrations on offer- most things are reporting via MQTT. It would make the integration of InfluxDB and Grafana more tricky plus any schedules, but I would not risk another of these episodes where something that works, is just suddenly removed. I’d run it on DietPi as it offers a reasonably good backup system.

4 Likes

The debate is aroud who has everything installed, configured, are just fine and don’t want to struggle with installing VM, hypervisor, backup things via docker cli, find docker repo, do cli things…etc etc.
HA is already complicated with its automations, connections and all, why should a person go to a learning curve such manage hypervisors, create VM, manage all of that, where they are just fine as they are now or they was with just a script on the OS they wanted to choose?
It’s not that hard to understand.
I don’t want do judge what you do in your house and in your envirnoment, you do your choices, so let’s other people do its choices.
The reason, i think, around many people did supervised on linux is because they just flashed the OS they was comfortable at, installed and runned with easy, launched a script and everything was there. Ready and with full power ( addons ) connected in one UI where you don’t necessary have to deal with docker things.
And without deal with hypervisor, VM, cli and all that stuff.
Deal that not everyone have time to invest on recreating an already running environment.
And if they do, it will be not as easy as it was before.
I get your point of view but it’s not everyone’s view :slight_smile:

HA future is going in user friendly direction-yaml is going away, integration in ui,… but with deprecating supervised on generic linux and to say solution is wm or core, that is a big step back in wrong direction. This step is not following the current concept of HA!

5 Likes

You were definitely doing something very wrong.

using a container manager like Portainer it is literally three clicks and you’re done. It takes a few seconds to pull the updated image and it re-creates the container immediately.

If you use the cli then it’s not much worse than that.

It usually takes way less than a minute or two from the time you start until you have the latest version running.

I never had any issues running my Unifi controller in a container at all. And updates were no big deal. The only reason I switched to a Cloud Key was because I thought the Unifi controller was screwy for HA presence detection. It turns out it was the HA Unifi Integration the entire time. :face_with_raised_eyebrow:

I’m very happy I went with Docker. Now I try to run everything I can in Docker containers.

4 Likes

too bad the poll already closed… :anguished:

1 Like

That’s OK

The point has been made I hope.

Correct. Losing Supervisor means losing all the tons of HA-addons, which provide the majority of it’s functionality! This is wrong indeed.

1 Like

looks at results Yep, the vast majority runs supervised in both polls. Point has been made! :smiley:

2 Likes

i too run the docker version with HassOS under generic Linux. Upgrading is nothing more than a point-and-click. Everything from there is automatic. Just wait for the reboot and it’s done.

Well… after a couple of day monitoring this channel, I’m wondering? Since supervisor generic is on hold and most probably wont be continue anyway. For those who own NUC why not run HASS core? Maybe someone here can create a simple docker-compose so it will be easier for everyone who are non tech skill to install? It will be almost as equal as supervisor right?

… or switch to Domoticz.
oops… :innocent:

1 Like

I do.

I have. Well, not really a docker compose but a good docker run command.

The problem is, that without supervisor, you have no add-ons. These make up for the most functionality of HA. I simply can’t afford to lose them. I would lose my panel in the living room, i would lose my password database, my ESP8266 driven devices, my monitoring, my influx HA-database, my SSH-access, my mobile app, all essential connectivity is provided by those plugins. We simply can’t do without.

5 Likes

I know… you are running docker core… since you share your code… hahahahaha… to bad VS Code doesnt install/run on RPI…

Again just like my explanations you actually need a docker-compose script that will assist you in installing the addon you would like to run. And boom just like supervisor you will have those addon running

Oh yeah, I forgot. Too many different posts to keep track of… :wink: