The web interface and background worker keep running, and tasks no longer fail after upgrades

Several things could stop the web interface or the background worker on a running device, or make tasks fail after every upgrade. Anthias 2026.06.3 fixes each of them.

Legacy time zone names

A device whose /etc/timezone holds an older name such as US/Central crash-looped the web interface and the worker at startup. These names work now, and the device keeps its local time: the images include the legacy time zone data. A name that still can't be found falls back to UTC.

Unreadable files on the SD card

If some of the web interface's own files on the storage couldn't be read, for example after file system damage, the web interface stopped at startup and restarted over and over. It now skips the files it can't read, starts, and logs one error per boot so the fault stays visible.

"Database is locked"

The web interface, the background worker and the viewer share one SQLite database. A write from one of them could make another fail at once with a database is locked error, for example when the viewer read its content list while the worker updated an item. They now wait up to 20 seconds for each other instead of failing, and reads no longer block writes. Databases from older installs switch over the first time they are opened. In this release the change also kept the viewer from noticing content you added or changed until it restarted. Anthias 2026.07.0 fixes that.

Errors after an upgrade or first boot

The worker starts alongside the web interface, which may still be updating the database after an upgrade or on first boot. Tasks that ran in that window failed with no such table: assets. The worker now waits until the database is ready, and logs a warning every 30 seconds while it does. Queued tasks stay queued and run once it starts.

On docker-compose installs, the web interface, worker and viewer also wait for Redis to answer before they start.

Reachability checks

The worker checks whether content that lives at a web address can still be reached. A check that ran past 30 seconds, for example on a slow DNS lookup, was killed along with the worker process running it. A check now gets 60 seconds, and one that runs out of time marks the content as unreachable. Content on your local network, such as an intranet page or another container on the same player, was still marked unreachable and skipped until Anthias 2026.07.0; until then, ANTHIAS_ALLOW_PRIVATE_FETCH=1 on the server and worker lets it through.