Why does Supervisor block a full restore without internet – wasn’t “local & offline” once HA’s core promise?
I deliberately run my HA system completely without internet access – my Zigbee sticks and connected devices are also intentionally denied any outbound connection, everything runs strictly local on my own network. That was one of the main reasons I chose Home Assistant in the first place: local control, no cloud dependency, no connectivity requirement – exactly what was advertised as a core principle of the project for years.
Now I’m trying to restore a full backup (all components checked, not a partial restore), and I get:
‘BackupManager.do_restore_partial’ blocked from execution, no supervisor internet connection
In plain terms: even a complete backup can’t be restored offline by Supervisor, because it internally checks whether internet is available – regardless of whether the add-on images are already present locally or not. To me this is a fundamental contradiction of the system’s whole premise: a backup that can’t be used offline in an emergency (hardware failure, migration) isn’t a usable backup for a system that’s deliberately run offline.
I’ve asked about this here before and got no factual answer, just comments along the lines of “refusing progress” and “incompetent” – with no mod stepping in. That doesn’t change the fact that the question is valid: why is a core feature like backup/restore tied to an internet check, when offline operation has explicitly been a supported and long-advertised way of using HA?
I’d appreciate factual answers – I already know the workarounds myself (disk image clone as a substitute), but I’d still like to understand the underlying design decision, especially from those who wrote the code.
There are just community members here, no devs. If you want an answer from devs, you have better chances asking on Discord than here on Discourse.
One thing it needs is the time, which it wants to get from an NTP server if you do not have a hardware RTC. I think it also wants to update libraries, because the bare install only contains the HA code and not the dependencies.
I also think the “local” refers to controlling devices, not the dependencies for updates of the software. But this is guesswork from me, I’m not affiliated with Nabu Casa.
In very basic terms a backup, backs up the the configuration files not the OS, integrations and apps. If you need a backup that can be restored without Internet you are going to need to make a copy of whatever storage medium you use.
This actually lines up with something I’ve been digging into. Before 2025.1, there was a clear Full vs. Partial backup choice, and a Full backup restored completely offline — I used it that way myself for years. Since the 2025.1 “complete overhaul” of the backup system (automated backups, retention policies, Cloud storage, extensible integrations), the distinction became “automatic” vs. “manual” instead, and that’s roughly the same point where people started reporting the 'BackupManager.do_restore_full/partial' blocked from execution, no supervisor internet connection error on restores that never needed internet before.
Doesn’t feel like a coincidence — the rework that added Cloud backup storage and integration-provided backup locations likely also introduced a blanket connectivity check that never differentiates whether internet is actually required for a given restore (e.g. a fully local, fully offline restore with everything already on disk).
And if that’s the case, why isn’t there an exception for people who deliberately run HA with no internet at all? Not everyone forgot to check a cable — some of us specifically chose HA because it didn’t require a connection, and now that exact use case gets punished by a check that was clearly built with Cloud/online users in mind.
Would be good to get confirmation from someone who worked on the 2025.1 backup rework on whether this was an intentional tradeoff or a side effect nobody caught.