Core won't update further than 2026.05.0

Installation
1

I have a notification for updating Core to 2026.05.3, but it's stuck. I click the update, the download starts, it installs, HA restarts and then the notification comes back again. I'm running in Docker which I know is unsupported blah, blah but has worked fine until now, so wondering what's changed/broken in 2026.05.0 that I need to workaround to get onto 2026.05.3

2

I'm on HAOS myself, but afaik you should download and deploy a new container, not try to update inside the old container. Assuming you mapped the data volumes outside docker that is, which you should always do.

3

I've been updating from within HA for five odd years with no issues. I've tried this time by re-creating the containers themselves and it's still not playing ball. I suspect something has gotten out of sync and this doesn't look right to me:

image
image578×177 6.7 KB

4

image
image610×148 6.36 KB

5

You are running HA supervised. That is deprecated. Another person had the same issue yesterday. False-positive automatic rollback on Core update (2025.8.0 → 2026.5.x) - Synology Supervised - Watchdog persists despite ignore-conditions

If you are already familiar with containers, I suggest switching to HA Container.

6

Alas the timeout was already set to 10 mins as that was fixed in a 2025 update.
It does look like the task of shifting to a Container install is now required.

7

This helped me yesterday: HomeAssistant update failed -> rollback! (2025.12.0) - #44 by lostprophet

I set it to 15 minutes and all have been good.

8

Thanks, I've tried setting to 15 mins and still stuck.

9

The logs may tell you whether it's respected or not. Try using AI to debug.

10

I've found the issue in the logs, but not sure how to fix. I tried a docker restart but unfortunately didn't help.

2026-05-20 20:00:06.269 ERROR (MainThread) [supervisor.docker.manager] Can't create container from ghcr.io/home-assistant/qemux86-64-homeassistant:2026.5.3: [400] invalid mount config for type "bind": bind source path does not exist: /run/supervisor

2026-05-20 20:00:06.270 CRITICAL (MainThread) [supervisor.homeassistant.core] HomeAssistant update failed -> rollback!

11

I was able to temporarily correct this by recreating the Supervisor container and binding the /run/supervisor volume while doing so. Core was updated (the labels and image details are 2026.5.4) and the update message was gone. But then Supervisor stepped in and 'updated' to 2026.5.0 with no interaction from me.

1 Like
12

Hi, wanted to chime in. I experienced the same issues going from 2026.4.1 -> 2026.5.1 it turned out i was using an ancient startupscript for the hassio-supervisor. Lots has changed, so to avoid other issues from popping up, i only updated the bind mounts.

Reference: https://github.com/home-assistant/supervised-installer/blob/c12badda6fcc7c70c3153a90241ba4da382671a5/homeassistant-supervised/usr/sbin/hassio-supervisor#L95

sudo cp /usr/sbin/hassio-supervisor /usr/sbin/hassio-supervisor.bak
sudo vim /usr/sbin/hassio-supervisor

Fix mounts, add missing stuff:

    -v /run/docker.sock:/run/docker.sock:rw \
    -v /run/containerd/containerd.sock:/run/containerd/containerd.sock:rw \
    -v /run/systemd-journal-gatewayd.sock:/run/systemd-journal-gatewayd.sock:rw \
    -v /run/dbus:/run/dbus:ro \
    -v /run/supervisor:/run/os:rw \
    -v /run/udev:/run/udev:ro \
    -v /etc/machine-id:/etc/machine-id:ro \

To apply:

docker stop hassio_supervisor
docker rm hassio_supervisor
systemctl start hassio-supervisor

NB: This is a ghetto fix, personally i wanna switch to a VM soon anyway. I expect zero longevity out of this.

Success:

INFO (MainThread) [supervisor.homeassistant.api] Connected to Core via Unix socket /run/os/core.sock

13

Thanks, unfortunately didn't work. After about 10 mins supervisor drops core back to 2026.5.0 again. I see in Portainer that all the additional volumes are missing again once the supervisor container is re-created.

14

I suspect the missing volumes are causing your issue, it uses the /run/supervisor mount to communicate with the Core (homeassistant) container. if no communication can be established within 10 minutes, it will automatically roll back.

Both the supervisor as well as the homeassistant container need to have these mountpoints.

homeassistant:

{
    "Type": "bind",
    "Source": "/run/supervisor",
    "Destination": "/run/supervisor",
    "Mode": "",
    "RW": true,
    "Propagation": "rprivate"
},

hassio_supervisor:

{
    "Type": "bind",
    "Source": "/run/supervisor",
    "Destination": "/run/os",
    "Mode": "rw",
    "RW": true,
    "Propagation": "rprivate"
},

Note: the 'Destination' (path inside the container) differs between the two.