About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
Not safe to delete by hand until you know what depends on it. In forward incremental, differential and log-backup designs an old full backup is the base for the incrementals, differentials or log backups after it, and removing the base makes them useless. Use the product's own delete operation, because it removes a whole chain and updates its records, then check copies, holds and the space you expect to free before you confirm. On deduplicated stores, deleting a snapshot may free no space until a later clean-up job runs.
What depends on an old full backup?
Incremental chains. Veeam states that to restore from a forward incremental backup you need a full backup file and a chain of subsequent incremental files. "If you delete a full backup file, the entire chain of incremental backup files will become useless." If you delete any incremental before the point you want, you cannot restore, because later incrementals depend on earlier ones.
That is why Veeam keeps a full longer than a naive count suggests. It does not create fulls automatically from retention limits, and "will remove the full backup chain only after the last incremental backup file in the chain becomes outdated". On some days there are more restore points on disk than the retention setting says.
Differentials. Microsoft documents that for SQL Server "before you restore a differential backup, you must restore its base". A differential holds only what changed since that base, so the base full is part of every restore that uses the differential.
Log backups. For SQL Server databases under the full or bulk-logged recovery model, a complete restore is a full backup, optionally a differential, then "all subsequent log backups (in sequence)". Microsoft also says the sequence of log backups is independent of data backups. Because the log backups are applied in sequence, one missing from the middle stops the roll-forward at that gap, even if every full still exists. Microsoft's page-restore text likewise calls for "an unbroken chain of log backups".
Shared data. In restic and Proxmox Backup Server, snapshots reference shared chunks. Removing one snapshot removes its record, not the data that other snapshots still use.
What will the tool do, and what will it refuse?
| Operation | What it does | Dependants |
|---|---|---|
| Veeam, Delete from disk | Permanently removes backup files from the repository and the console; deletes the whole chain | The next job run creates a new full for the affected VMs |
| Veeam, deleting files by hand | Not recommended | Veeam says subsequent backup or replication sessions will fail |
| Veeam, Remove from configuration | Removes the records only; the files stay in the repository | The backup can be imported again later |
| Veeam, Forget or Remove from disk on missing restore points | Available only for points missing from the chain or those depending on them | The option is not offered for a healthy point |
restic forget then prune |
forget removes snapshots; prune removes data only they used |
--dry-run shows the plan without removing anything |
Proxmox Backup Server prune or snapshot forget |
Removes only the backup index files, not the data | A snapshot with the protected flag is not pruned or manually deleted |
| Proxmox Backup Server garbage collection | Removes unused chunks, with a grace period | Chunks used by any snapshot are kept |
What should you check before deleting?
- Dependants. List what restores from this full: incrementals, differentials, log backups, synthetic fulls. In Veeam, use Delete from disk, which deletes the whole chain, instead of removing one file.
- A second copy. Check where the other copies are. In Veeam, "Delete from disk" on a scale-out backup repository "will remove the backups not only from the performance tier but also from the capacity and archive tier". For backup copy jobs, the same operation removes copies from all tiers, including detached backups shown as Orphaned.
- Holds and obligations. Check whether the data is subject to a legal hold, an audit request or a records rule. Check these before any delete; the related piece on holds covers them.
- GFS flags. In Veeam, the Delete from disk page says to select the "Remove GFS full backups" check box "to remove backups with GFS flags (weekly, monthly and yearly)". Read the dialog and decide deliberately whether to tick it.
- Encryption keys. Removing an encrypted backup from the configuration removes its keys from the database, so importing it later needs the password.
- Space. On deduplicated stores, forgetting the snapshot does not free space at once. Proxmox's documentation says neither prune nor forget frees space "in the chunk-store"; garbage collection does that, and the sweep keeps chunks within a grace period of 24 hours and 5 minutes.
- A restore test of what remains. After deleting, restore a point from the surviving set.
What does this look like in numbers?
Example (illustrative): fs01 is backed up nightly to backup01, with a full every Sunday, retention of 14 restore points and Veeam-style forward incrementals. Number the days 1 to 21, with Sundays on days 1, 8 and 15, so there are three chains: A (days 1 to 7), B (days 8 to 14), C (days 15 to 21).
On day 14 the retention window is days 1 to 14. All of chain A is inside it, and it needs its full on day 1 for every point from day 2 to 7. Someone sees the old full from day 1, now 13 days old, and deletes that file by hand to save space. Seven restore points (days 1 to 7) are lost, which is half of the 14 the policy promised. Veeam also states that manual file deletion makes subsequent jobs fail.
If the tool does the work, it waits. On day 20 the window is days 7 to 20, and day 7 is the last point of chain A, so chain A stays: 20 points are on disk against a retention of 14. On day 21 the window is days 8 to 21 and chain A can go, leaving 14 points. The extra 6 points on day 20 (20 minus 14) are the price of safety.
What if the real problem is space?
If you are thinking of deleting an old full backup because the storage is filling, stop first. Deleting one file is the riskiest way to fix that. Safer options, in order:
Find what is using the space. A chain that is waiting for its last incremental to age out, a detached or orphaned backup, or a store that has not yet run its clean-up can all hold space that looks wasted.
Shorten the retention in the product's settings, so the tool removes whole chains in the right order, instead of removing a single file.
Run the product's clean-up and check the result. On Proxmox Backup Server the documentation recommends scheduling garbage collection regularly, with a weekly schedule as a good starting interval.
Disruptive step: shortening retention makes the tool delete the older chains, and the clean-up then removes their data for good. Restore points older than the new setting cannot be recovered except from a copy held elsewhere. Write down the old setting first; putting it back protects only the points that still exist.
Add storage or move older copies to a second location, then confirm they restore.
Whatever you choose, record what you removed, when, and which restore points no longer exist. That record answers the question an auditor or a lawyer will ask later.
What does the delete command look like on each system?
These are the documented forms. Run the read-only step first.
restic. restic forget --keep-daily 5 --dry-run prints what would be removed and removes nothing. Then, when you are sure: Disruptive step: restic forget <snapshot ID> removes the snapshot, and restic prune removes the data nothing else uses. Pruning locks the repository for its duration, so backups cannot complete. Neither command can be undone. restic advises running restic check after pruning.
Proxmox Backup Server. proxmox-backup-client prune <group> --keep-daily 7 --keep-weekly 4 --keep-monthly 3 with --dry-run first, which "only shows the list of existing snapshots and what actions prune would take". To protect one snapshot from pruning or deletion: proxmox-backup-client snapshot protected update <snapshot> true. Disruptive step: proxmox-backup-client snapshot forget <snapshot> and proxmox-backup-client garbage-collect cannot be undone.
Veeam. In the Home view, select Backups, choose the backup, then Delete from > Disk. Disruptive step: this deletes the whole chain permanently. You can instead use Remove from > Configuration, which keeps the files and removes the records. For that operation Veeam advises creating an encrypted configuration backup first.
How do you check your own system or provider?
Ask in writing:
- If I remove this restore point, which other restore points can no longer be restored?
- Does the delete go through the product's own operation, and does it remove the whole chain?
- Where are the other copies, and does deleting here also delete there?
- Are any points flagged for weekly, monthly or yearly retention, or under a legal hold?
- How and when is space returned after a delete?
- Can a deleted restore point be recovered, and for how long?
Read next
Related articles you may want next.
- Backup & RecoveryRetention that outlasts your ability to restoreHow retention periods, backup chains, keys and obsolete formats come apart, and how to test the path you claim to keep.Read article
- Backup & RecoveryGrandfather-father-son retention: counting the restore points a policy really keepsHow to count the restore points a 30 daily, 12 monthly backup policy keeps, where the gaps fall, and what to ask a provider to confirm in writing.Read article
- Backup & RecoveryLitigation holds and backups: when old backups become evidenceHow Australian discovery treats backups, why routine deletion matters once litigation is anticipated, and how to plan a hold without misusing backups.Read article
Risks and limits
The file-level counts in the example follow the documented Veeam rule for forward incrementals, and real products differ in detail. Microsoft's pages describe SQL Server restore sequences, not every backup product's catalogue. Deletes are permanent unless another copy exists, so take the check list as a floor.
These commands have not been run or lab-tested.
NetLeverage includes a daily full disaster-recovery backup of each hosted server at no extra charge. Talk to an engineer.
Sources
- Forward Incremental Backup Retention Policy — Veeam, page updated 2026-01-12, accessed 17 November 2026
- Deleting Backups from Disk — Veeam, page updated 2025-02-10, accessed 17 November 2026
- Removing Backups from Configuration — Veeam, page updated 2025-07-09, accessed 17 November 2026
- Removing Missing Restore Points — Veeam, page updated 2026-02-05, accessed 17 November 2026
- Removing Backups (backup copy) — Veeam, page updated 2026-07-23, accessed 17 November 2026
- Differential backups (SQL Server) — Microsoft Learn, updated_at 2026-10-19 (ms.date 2025-07-15), accessed 17 November 2026
- Restore and recovery overview (SQL Server) — Microsoft Learn, updated_at 2026-10-19 (ms.date 2024-12-04), accessed 17 November 2026
- Transaction log backups (SQL Server) — Microsoft Learn, updated_at 2026-10-19 (ms.date 2024-07-08), accessed 17 November 2026
- Removing backup snapshots — restic project, date not shown, accessed 17 November 2026
- Maintenance Tasks — Proxmox Server Solutions, date not shown, accessed 17 November 2026
- Backup Client Usage — Proxmox Server Solutions, date not shown, accessed 17 November 2026
This article was originally published on .