Restore practice: file-level, VM-level, restore certification

Why this matters in production

Backups that have never been restored are not backups. They are hope. This lesson teaches the restore operations and the restore-certification drill.

File-level restore

PBS can mount individual backups and restore specific files:

proxmox-file-restore mount <backup-id> /mnt/restore

Browse the mount, copy the needed files, unmount:

umount /mnt/restore

GUI: Backup → select snapshot → File Restore. Provides a tree view of the backup.

Full VM restore

GUI: Backup → select snapshot → Restore → choose target storage.

CLI:

qmrestore <backup-archive> <new-vmid> --storage local-zfs

The new VM has a different VMID; it does not collide with the original. Verify it works before deleting the original.

Restore certification drill

A monthly exercise that proves the restore path works end-to-end:

1. Pick a random VM (production, but not too critical).
2. Restore it to a fresh VMID in an isolated network.
3. Boot the restored VM.
4. Run a verification script inside the VM (e.g., check that key services start).
5. Document the result.

flowchart LR
  A[Pick random VM] --> B[Restore to isolated network]
  B --> C[Boot restored VM]
  C --> D[Run verification script]
  D --> E[All checks pass?]
  E -->|yes| F[Restore certificate: pass]
  E -->|no| G[Restore certificate: fail]
  G --> H[Investigate, fix, repeat]

Restoring into an isolated network

The restored VM must be on a network that does not conflict with the production VM. Use a dedicated SDN VNet or VLAN for the restore-certification lab.

flowchart TB
  subgraph PROD[Production network]
    P1[VM 100]
    P2[VM 101]
  end
  subgraph LAB[Restore lab network]
    R1[Restored VM 200]
  end
  PBS[(PBS)] -->|restore| LAB

Restoring from off-site PBS

When the primary site is gone, restore from the off-site PBS:

proxmox-backup-client restore <host> <backup-id> <snapshot> <target-path>

For full VM restore, attach the remote PBS as a storage entry and restore from it.

Restore documentation

A complete restore procedure includes:

• Who has the authority to declare a disaster and trigger a restore.
• Where backups live (local PBS, off-site PBS, S3, tape).
• How to attach off-site storage when needed.
• The sequence in which systems are restored (covered in the DR lesson).
• How to verify each restored VM.
• How to communicate status to stakeholders.

A break/fix exercise

Production considerations

Common mistakes

• “The backups look fine, I don’t need to restore-test.”
• Restoring over the production VM.
• Not testing off-site restore.

Key takeaways

• File-level restore is fast and works on running VMs.
• Always restore to a new VMID first.
• Run restore-certification drills monthly.

Knowledge check