Quick answer: Retention is how long each backup is kept before deletion. The classic grandfather-father-son scheme balances history against disk: keep 7 dailies, 4 weeklies and 6–12 monthlies. It protects you from problems noticed late — corruption or a hack found weeks after it happened.
Overview
Frequency answers "how fresh is my newest backup"; retention answers "how far back can I go". Ransomware that encrypted files three weeks ago, a bad migration noticed after a month, a deleted record a customer asks about in a quarter — all are retention problems, and a single overwritten mirror fails every one of them.
Before you start
- Know your backup frequency (see how often to back up).
- Know any legal/contractual minimums for your data (invoices, personal data, logs).
- Storage math: retention multiplies copies — incremental tools make long retention affordable.
Step-by-step guide
- Pick tiers — a common GFS baseline: dailies kept 7 days ("son"), weeklies kept a month ("father"), monthlies kept 6–12 months ("grandfather").
- Implement with dated directories and
find -mtimepruning, or let a snapshot tool do it:restic forget --keep-daily 7 --keep-weekly 4 --keep-monthly 12 --pruneborg prune --keep-daily=7 --keep-weekly=4 --keep-monthly=12 - Make pruning part of the same scheduled job as the backup, and log it.
- Review quarterly: does the oldest backup still cover your "noticed late" scenarios?
Common issues
- Mirror-only "retention": rsync with
--deleteand no snapshots keeps exactly one version — yesterday's mistake is already everywhere. - Retention without pruning: keeping everything fills the disk and then backups stop; deletion must be automated too.
- Pruning before verifying: never let a prune run delete the only good copy — verify first (see integrity testing).
When to contact support
If your retention needs outgrow your Storage VPS plan, open a ticket — upgrading disk space or adding a second target in another location is straightforward.
Frequently asked questions
What is grandfather-father-son (GFS) retention?
A tiered scheme keeping recent history dense and older history sparse — typically 7 daily, 4 weekly and 6–12 monthly backups, which covers late-discovered problems affordably.
Why keep old backups if I have last night's copy?
Because problems are often noticed late: corruption, a bad migration or an intrusion from three weeks ago is already inside every recent backup — only older snapshots predate it.
How do restic and borg handle retention?
Both have built-in policies: "restic forget --keep-daily 7 --keep-weekly 4 --keep-monthly 12 --prune" or the equivalent "borg prune" enforce GFS automatically after each run.
Related articles
- How often should you back up your server?
- How to estimate backup storage size
- How to test backup integrity
- Common backup mistakes
Need reliable space for your backups? Order a Storage VPS at Cloud2Y →
