Direct answer
Retention means keeping more than the newest backup. Use multiple recovery points so a corrupted world, broken plugin migration or accidental deletion is not silently propagated into every copy.
Design the retention tiers
- Set a recovery point objective and required history: short intervals for recent accidents and longer retention for issues discovered days later.
- Example policy for an active server: a limited window of hourly snapshots, several daily copies and a few weekly or monthly copies. Adjust to available storage and legal/privacy requirements.
- Tag pre-upgrade backups separately; do not let automatic rotation immediately delete the only known-good rollback.
- Calculate how much full archives and any incremental chains consume; an incremental chain is useless if its required base has been deleted.
Avoid common mistakes
Retention is not the same as replication: putting all backups on the same disk leaves them exposed to disk failure. Never configure a deletion job until you have verified which directory it targets, and do not include the current live world in a cleanup glob.
Version and safety notes
Retention and deletion scripts must be tested on non-production copies. Keep dated recovery points off-host and verify that incremental bases are retained.
Verify recovery
Maintain a dated manifest containing the Minecraft/software version, worlds, plugin data, size and test-restore status. Periodically restore an older backup, not just the newest one, to a separate location. Confirm database and world state agree with the expected point in time.
References
Check documentation for the versions and hosting environment you actually run.
Related guides
Help us improve this article
Tell us what is missing, unclear, outdated, or incorrect. You can also suggest something we should add.