Backups and restores
When backups run, how long they are kept, and how to restore without losing the last hour of play.
Schedule
The backup worker runs between 02:00 and 05:00 local node time. The instance is stopped for the duration of the copy, which is usually under thirty seconds for a world up to a few gigabytes. If your group plays at night, set a different window per instance in Backups → Schedule.
| Plan | Frequency | Retention |
|---|---|---|
| Sandbox 2 | Daily | 7 days |
| Squad 4 | Daily | 14 days |
| Community 8 | Daily or hourly | 21 days |
| Dedicated node | Configurable | 30 days |
Restoring
- Open BackupsThe list shows date, size and a checksum prefix for each snapshot.
- Pick a snapshotRestore always targets the live instance. If you want to inspect a backup first, use “Restore to a new instance” instead.
- Confirm the stopThe panel stops the server before restoring. It refuses to run if the instance is mid-save.
- Wait for the verifyA restore is copied, checksummed and then swapped in. Typical time is one to three minutes for a 4 GB world.
Moving a backup elsewhere
Snapshots are plain directories, not a proprietary format. You can download one over SFTP from /backups and open it locally with the same game server. This is also the recommended way to keep an off-site copy if you do not want to rely on us — please do.
What is not in the backup
- The game binaries and the mod jars (they are re-downloaded or re-imported).
- The console logs under
/logs. - Anything you placed outside the instance directory.
If a world fails to load after an ungraceful shutdown, restore the most recent backup rather than trying to repair region files by hand. Restoring is faster and does not risk making the damage worse. Then look at the console: the crash usually happened for a reason worth fixing.
Retention changes in 2026
As of panel 2.9.0, retention is set per plan rather than per node, so a plan change no longer silently changes how far back you can restore. Existing snapshots are unaffected.