Environment
- Host: Synology NAS (DSM)
- Installation: Home Assistant Supervised via Docker
- Supervisor: 2026.05.0
- Core: Stuck at 2025.8.0 (attempting to upgrade to 2026.5.3)
- Stability: This setup has run flawlessly for ~3 years with zero upgrade issues until this version jump 2 days ago.
The Problem
The Supervisor automatically rolls back the Core update exactly ~10 minutes after starting the new container. The Core actually starts successfully, and the UI is fully accessible on
the new version.
The rollback is triggered because the Supervisor fails to receive the RUNNING state signal in time. Crucially, this happens even when I have explicitly set the Supervisor to ignore
health conditions, which suggests the internal 10-minute update watchdog is operating independently of the health check flag.
Logs (Supervisor)
Here is the timeline of a recent attempt showing the rollback occurring despite the ignore-conditions flag:
10:31:02 INFO [supervisor.homeassistant.core] Updating Home Assistant to version 2026.5.3
10:31:02 CRITICAL [supervisor.jobs] The following job conditions are ignored and will make the system unstable when they occur: {<JobCondition.HEALTHY: 'healthy'>}
...
10:31:32 INFO [supervisor.docker.homeassistant] Starting Home Assistant ... version 2026.5.3
[... 10 minutes pass, UI is fully responsive on 2026.5.3 ...]
10:41:33 ERROR [supervisor.homeassistant.core] No Home Assistant Core response, assuming a fatal startup error
10:41:33 CRITICAL [supervisor.homeassistant.core] HomeAssistant update failed -> rollback!
10:41:33 INFO [supervisor.docker.interface] Updating image ...:2026.5.3 to ...:2025.8.0
What I've Tried
- Ignore Conditions Set: I ran ha jobs options --ignore-conditions healthy. The Supervisor logs confirm this flag is active (CRITICAL ... ignored conditions), yet the rollback still
triggers at the 10-minute mark. - Missing Mounts: I resolved the missing /run/supervisor directory issue (which initially prevented container creation).
- No Crash: The Core does not crash. It stays running and accessible via the API during the entire 10-minute window before the Supervisor forcibly stops it to perform the rollback.
Question
The ignore-conditions healthy flag seems to bypass the pre-update health check, but the Supervisor's internal 10-minute update watchdog is still aggressively killing the process
because it hasn't received a "Ready" signal.
Is there a way to extend or disable this specific 10-minute watchdog timer for the Core container update? Or how can I manually signal to the Supervisor API that the Core is ready so
it stops forcing this rollback?
