Proxmox VE guide (5): backups and snapshots (and why they're not the same thing)
Snapshots for experimenting, vzdump backups and their three modes, scheduled jobs with retention, restoring and when Proxmox Backup Server is worth it.
Where this guide comes from: at work I administer a Proxmox VE cluster every day, with Ceph storage and backups to Proxmox Backup Server. This series is the version for starting from scratch at home, with a single server, and follows the official Proxmox VE 9.2 documentation. And at home I once lost three nights of backups because I didn’t notice they were failing, so this topic is close to my heart.
We’ve reached the most important part of the guide. Having lots of machines on one server is wonderfully convenient… until that server fails. Let’s protect it.
Snapshots: for experimenting
A snapshot freezes a machine’s state at a point in time. If you break something afterwards, you go back to that point in seconds.
- Machine → Snapshots → Take Snapshot. You can include RAM, to return with programs open exactly as they were.
- Rollback to go back.
Use them before every risky change: an update, a config change, a class exercise.
But they’re not backups
Snapshots are stored on the same storage as the machine. If that disk dies, the machine and all its snapshots go together. Also, piling up lots of them over months slows things down and takes space. They’re an undo button, not insurance.
Backups with vzdump
Proxmox includes its own backup tool, vzdump. It creates a file with the complete machine (configuration and disks) that you can restore on this or another server.
The three modes
| Mode | What it does | Downtime |
|---|---|---|
| snapshot | Copies the machine while it keeps running | None |
| suspend | Pauses it briefly to copy it | Short |
| stop | Shuts it down, copies it and starts it again | Longest, but the cleanest copy |
For virtual machines, snapshot is the usual choice: the VM keeps running and, with the QEMU agent installed (see part 3), Proxmox freezes the filesystem for an instant so the backup is consistent. For LXC containers, snapshot mode needs storage that supports snapshots (LVM-thin, ZFS…).
Where backups go
By default, to the local storage. A bad idea as your only copy: it’s on the same disk as the machines. Better to add another destination:
- Datacenter → Storage → Add → SMB/CIFS or NFS, pointing for example at a folder on your NAS.
- Under Content, tick Backup (some versions show it as “VZDump backup file”).
Scheduled jobs
Datacenter → Backup → Add:
- Storage: the NAS one.
- Schedule: for example, every day at 2:00.
- Selection mode: all machines, or just some.
- Mode: snapshot.
- Compression: ZSTD (fast, compresses well).
- Retention: how many copies to keep. For example,
keep-daily=7,keep-weekly=4,keep-monthly=3. Proxmox prunes the extras automatically. - Notification: have it email you if anything fails. Don’t skip this.
Restoring
Select the backup storage → Backups → pick one → Restore. You can restore over the original machine or as a new machine with a different ID, which is perfect for checking the backup works without touching the original.
A backup you’ve never restored is a hope, not a backup. Test restoring a machine now and then.
What about Proxmox Backup Server?
Once you have several machines and lots of backups, vzdump falls short: each backup is a full file. Proxmox Backup Server (PBS) is a separate product, also free, built for this:
- Deduplication: data repeated across backups is stored only once. It takes far less space.
- Incremental backups: after the first, it only copies what changed.
- Automatic verification that backups are still readable.
Ideally install it on another machine (it makes little sense to store backups on the same server they protect). For a small lab, vzdump to a NAS is enough; for something more serious, PBS.
The 3-2-1 rule
The classic backup rule:
- 3 copies of your data (the original and two backups).
- On 2 different media (for example, the server’s disk and the NAS).
- 1 off-site (a disk at a relative’s home, or in the cloud).
That way, a single disaster (a disk, a theft, a fire) doesn’t take everything.
Guide summary
- Install with a fixed IP and a network cable.
- Repositories without subscription, update with
full-upgrade, your own user with two-factor auth. - Virtual machines with VirtIO and the QEMU agent.
- Unprivileged LXC containers for simple Linux services.
- Scheduled backups off the server, with alerts, and test restores.
If you’ve made it here reading all five parts, there should be a new achievement waiting for you on the achievements page.
Did it stick?
Three quick questions. Each right answer is worth 10 XP.