Backup and restore
- Backup & Restore: Enabled by default; the installer renders the archive PVC and its path.
- Manual backups: archive a server's entire data volume (
/data, including worlds, configuration, plugins and mods) to the cluster's archive store, with rollback to any backup point - Daily restore points: a server played that day gets a restore point once it stops; by default 7 are kept for up to 90 days, rotated separately from manual backups
- Download and export: download a single backup (with sha256 verification), delete a single backup, or export a whole world
- Off-site copy (optional): backups are encrypted on the host and synced to S3-compatible storage (AWS S3, Cloudflare R2, Backblaze B2, MinIO and others)
- Control-plane database: the database holding accounts, server ownership, quotas and the archive index is backed up daily and snapshotted before every upgrade migration;
felis db restorerolls it back atomically, and the panel's Maintenance & Backups page shows the age of the latest backup (see troubleshooting §16)
- Manual backups: archive a server's entire data volume (
What a backup contains
A backup tars the server's ENTIRE data volume — the same volume the server mounts at /data: world folders, server.properties, plugins/mods, configs, jars, libraries, logs and cache, not just the world/ directory. A restore replaces the volume's contents with the archive (files added since the backup are pruned), so a restore also rolls config/plugin changes back. Sizes are dominated by libraries/cache on stock Paper servers (~170MB for a fresh instance before any world growth) — do not size the archive PVC as if only world data were stored.
Detailed procedures
- World backups, downloads, exports and daily restore points
- Control-plane database backups and disaster recovery
- Disaster recovery and planned migration
Source: README_EN.md, docs/troubleshooting.md, docs/operations.md.