3-2-1 Backup Strategy with Proxmox Backup Server

Backups are like fire extinguishers: invisible for years, but in an emergency, the only thing that matters. With Proxmox Backup Server (PBS), we have a solution that does more than nightly full backups to external drives.

3-2-1: The Standard That Applies

• Three copies of data (original + two backups)
• On two different media (e.g., NVMe storage + HDD array)
• One copy offsite (different fire zone, different building, different city)

PBS covers this out of the box — and NIS2 effectively demands nothing else.

From 3-2-1 to 3-2-1-1-0

Since ransomware now specifically targets backups for encryption, 3-2-1 alone is no longer sufficient. The extended rule: 1 copy offline or immutable, 0 errors after verification. This is exactly where PBS shines — more on verification jobs and an append-only offsite copy below.

Proxmox Backup Server: Deduplication as a Killer Feature

PBS backs up VMs and containers incrementally with block-level deduplication. The initial run takes time; every subsequent run backs up only changed blocks. The result: hundreds of snapshots using surprisingly little storage.

A PBS backup of an ERP VM with a 200 GB disk occupies barely more than 250 GB after 90 days of retention — because only a few hundred MB change daily, and the rest is referenced.

proxmox-backup-client backup root.pxar:/ --repository senn@pbs@<pbs-ip>:backups

The repository can be local — or remote. Exactly what we use for the offsite copy.

Verify, Prune, Garbage Collection

Three routines keep the datastore healthy:

• Verify jobs read back the chunks and check their checksums — making the "0" in 3-2-1-1-0 reliable, not just hoped for.
• Prune deletes according to a retention policy (e.g., "7 daily, 4 weekly, 6 monthly").
• Garbage collection frees storage from chunks no longer referenced.

In larger environments, namespaces cleanly separate tenants or locations within the same datastore.

Offsite: Hetzner Storage Box as PBS Remote

The Hetzner Storage Box (not in the same data center as the cluster) is integrated as a PBS remote. After the local backup, PBS replicates the snapshots there:

proxmox-backup-manager remote add hetzner-offsite \
  --host <storagebox-ip> --user senn --repository backups

This is not a manual sync, but PBS-native replication — encrypted, deduplicated, incremental. With restricted write permissions on the remote side (append-only), this becomes the immutable copy that even a compromised cluster cannot reach. The Storage Box costs less than a cloud backup subscription and keeps data in Germany.

ZFS Snapshots as the First Line of Defense

Backup is the last resort; snapshots are the first. When something goes wrong — a faulty SQL update, a deleted share — the snapshot from three hours ago is restored faster than any tape.

PBS integrates with ZFS: create snapshot, back up snapshot, discard snapshot. No running VM freezes, no service notices. Even the TrueNAS file server hooks into this backup chain.

And If VMware Is Still in the Picture?

In mixed environments, Veeam Backup & Replication (since version 12.2 with native Proxmox VE support) backs up VMware, Proxmox, and physical systems from a single console — and can restore a VMware backup directly to Proxmox. This is also an elegant migration path away from VMware.

Restoration: Tested, Not Believed

A backup never restored is no backup. We monthly restore a test VM from the offsite copy. PBS doesn’t automate this, but it keeps us honest. Duration: under 15 minutes, including boot test.

Conclusion

3-2-1 sounds like theory; PBS makes it practical. Local backup on the cluster, immutable offsite copy on Hetzner Storage Box, snapshots for quick recovery, verify jobs for confidence — this is the backup strategy that even convinces a NIS2 auditor.