Good way to rollback supervisor on supervised installations?

Hi,

has anyone attempted to rollback the supervisor container to a specific version and keep it there?

I have a system that recently became unhealthy due to an unupgradable Docker 18.x, outdated networkmanager, etc… so I wanted to revert the supervisor to a version that worked before, so I rolled it back from 2020.12.x to 2020.11.2 to see what happens.

Has anyone done something like this with success?

Below are my steps.

While these generally worked out I’m aware that this is just a dirty hack to see how the system reacts.

There is still the update notification to 2020.12.06 and a log entry about a failed update attempt. I blocked the update URL with pihole, but apparently requests are still getting out despite having hassio DNS configured to use pihole.
The unhealthy message in the status is gone but it is still in the logs.
The connectivity check seems to fail as well (probably due to pihole this time)

I’m wondering whether something like this could work at all.
And what the implication are for addons and their updates…




I stopped the supervisor service, fiddled with the hassio.config and the start script:

sudo systemctl stop hassio-supervisor.service
cat /etc/hassio.json
{
   "supervisor": "homeassistant/armhf-hassio-supervisor:2020.11.2",
   "homeassistant": "homeassistant/raspberrypi3-homeassistant",
   "data": "/usr/share/hassio"
}

sudo nano /usr/sbin/hassio-supervisor
<added this to docker run command>      
        --privileged \
        --security-opt apparmor=unconfined \

Removed the old container and image. Pulled the correct one (2020.11.2 in this case).
And restarted the service again

sudo systemctl start hassio-supervisor.service

1 Like

Danke Jochen,
you just saved me a lot of headaches with the description of your solution!
After update to Supervisor 2022.06.2 Home Assistant did not start anymore. Following your steps I could revert back to 2022.05.03 and it’s running again.