How to rollback to an older snapshot without deleting newer ones?

[Perfeitor]

Hi everyone,

I'm currently using cv4pve-autosnap on Proxmox to automatically create scheduled snapshots, with labels or tags by week for easier tracking.
The storage backend is LVM Thin, which I chose to avoid full disk usage when running VMs. This setup works well for my needs, and I don’t want to change it.

However, I’ve run into a frustrating issue.

---

The issue:
When I try to rollback to an older snapshot (for example, one from 3 days ago), Proxmox requires me to delete all newer snapshots first before allowing the rollback.

This is a big limitation. I sometimes need to revert to a previous state to troubleshoot issues or verify behavior, but I still want to keep the newer snapshots for comparison or as independent backups.


This behavior occurs with both VMs and LXC containers.
(The screenshot I attached below is only to illustrate the snapshot structure — not tied to a specific VM/LXC.)

---

Things I’ve already tried (unsuccessfully):

1. Clone from the desired snapshot
   → Doesn't work. Proxmox throws an error due to the VM having an efidisk0.
2. Backup the current state, delete the newer snapshots, and then rollback
   → Even worse. When restoring from the backup, all snapshots are lost, since Proxmox backups don’t preserve snapshot history.

---

What I’m looking for:
Is there any way to roll back to an older snapshot without having to delete the newer ones?

Or is there a workaround that lets me temporarily restore or test an older state, while still keeping the entire snapshot chain intact?

Any suggestions would be greatly appreciated. Thanks in advance!

 

[_gabriel]

Snapshots are not backups.

Proxmox Backup Server is made for daily, even more, backups, as only "différentiel" data are saved on the backup drive, like your daily snapshot, then to test older backups, restore wanted backup as a new VM.

 

[fiona]

Hi,
are you really sure that ZFS is not involved? The limitation with only being able to rollback to the latest snapshot should only apply to ZFS, not LVM-Thin.
Please share the container configuration pct config 107, the storage configuration cat /etc/pve/storage.cfg and output of pveversion -v. Also the VM configuration for the issue with the EFI disk, qm config <ID> and the exact error message there.

 

[UdoB]

The limitation with only being able to rollback to the latest snapshot should only apply to ZFS

Is there any progress in to utilize "zfs clone" (linked or full/promoted) to create this missed so sadly feature? Each and every snapshot should be usable to spawn a new instance. Possibly with a new <vmid> to stay conflict-free...

Just dreaming... ;-)

 

[fiona]

Hi,

Is there any progress in to utilize "zfs clone" (linked or full/promoted) to create this missed so sadly feature? Each and every snapshot should be usable to spawn a new instance. Possibly with a new <vmid> to stay conflict-free...

Just dreaming... ;-)

AFAIK, nobody is currently working on this, tracked here: https://bugzilla.proxmox.com/show_bug.cgi?id=2805

 

[Perfeitor]

Hi,
are you really sure that ZFS is not involved? The limitation with only being able to rollback to the latest snapshot should only apply to ZFS, not LVM-Thin.
Please share the container configuration pct config 107, the storage configuration cat /etc/pve/storage.cfg and output of pveversion -v. Also the VM configuration for the issue with the EFI disk, qm config <ID> and the exact error message there.

Yes, I am indeed using ZFS. Therefore, as you mentioned, the inability to roll back to older snapshots is expected behavior with ZFS.

I’ve also considered doing full backups instead of relying on snapshots. However:

• With the current snapshot frequency (e.g. every 4–6 hours), full backups would require hundreds of gigabytes per week per VM.
• In addition, backups take significantly more time to complete compared to snapshots.

Meanwhile, ZFS snapshots are extremely space-efficient — I can retain the full state of a VM across an entire week, and all the snapshots combined consume less than 1GB, which is simply not feasible with full backups.

If there's any alternative approach that allows retaining multiple restore points without such a heavy storage cost, I’d greatly appreciate your advice.

 

[_gabriel]

• With the current snapshot frequency (e.g. every 4–6 hours), full backups would require hundreds of gigabytes per week per VM.
• In addition, backups take significantly more time to complete compared to snapshots.

Meanwhile, ZFS snapshots are extremely space-efficient — I can retain the full state of a VM across an entire week, and all the snapshots combined consume less than 1GB, which is simply not feasible with full backups.

This is why Proxmox Backup Server exist.
Backups are all full, but data are always dedup = same space efficiency as snapshots.
Backups are really fast after the first backup, thanks to qemu dirty bitmap which allow read and backup only changed data blocks since previous backup, but from point of view of Proxmox Backup Server, all backups are full, but remember same data is only stored once on the backup disk.

 

[Perfeitor]

This is why Proxmox Backup Server exist.
Backups are all full, but data are always dedup = same space efficiency as snapshots.
Backups are really fast after the first backup, thanks to qemu dirty bitmap which allow read and backup only changed data blocks since previous backup, but from point of view of Proxmox Backup Server, all backups are full, but remember same data is only stored once on the backup disk.

Thanks a lot for the clarification! I didn’t realize PBS handled deduplication so efficiently — I’ll definitely look into it. Appreciate your help!