Render Queue Worker
ActiveFilesystem-based render job queue for a Linux render node: atomic claims, crash recovery, systemd supervision.
A small daemon that turns a watched directory into a render job queue. Jobs are JSON
files; claiming a job is an atomic mv; recovery after a crash means moving files
back. No database, no broker, no web UI.
Design notes
- The filesystem is the queue —
mvgives atomicity for free. - systemd handles restarts, logging and memory limits.
- Notifications go out through a Telegram bot on job completion or failure.
Status
Running in production on my render node since mid-2026. Next step: a --dry-run
mode that estimates job duration from scene statistics.