Veeam vs Proxmox Backup Server
on Proxmox VE

Two products, two models

Veeam Backup & Replication is a commercial platform with a console, repositories, workers and per-hypervisor plug-ins. PBS is a separate AGPL server with a client in PVE. We do not invent Veeam list prices here — they are quote / per-socket in a Veeam contract. PBS subscription pricing is public at Proxmox (the same model as VE: Enterprise repo + portal from Basic up). That is why this article compares feature scope and limits, not “Veeam vs €550”. A Veeam quote goes through a partner; PBS runs without a subscription, the Enterprise repo is optional.

Veeam plug-in limits

Current Veeam helpcenter (build 13.x): the plug-in supports Proxmox VE 8.2–9.2 from the Proxmox ISO, x86-64 (not ARM). It needs at least one file-level storage in PVE. Application-aware and item-level restore need the QEMU Guest Agent in the VM, enabled in PVE, before backups are taken.

Official limitations (Considerations and Limitations):

• No LXC container backup.
• No VM template backup.
• No VMs that share a BIOS UUID.
• iSCSI disks attached to VMs and passthrough disks are skipped.
• Btrfs and custom storage — no. Other PVE types: yes. LVM with SAVC: yes.
• Restore of VM HA settings — no.
• Restore from tape straight to PVE — no (repository first).

Concurrent backups per storage: 4 (change via Support). Instant Recovery to a PVE host is experimental. Do not customise workers unless Support asks.

What PBS does natively

PBS: VMs and CTs, chunk dedup, incremental, Zstandard, client-side AES-256-GCM, SHA-256, sync, tape, namespaces, Datastore.Backup without delete. Broader service write-up: Ransomware-resistant backup. There is no Veeam-style VSS application-aware — Windows consistency goes through the QEMU guest agent / fsfreeze in the PVE job, not Veeam Application-Aware Processing.

Comparison table

When Veeam, when PBS, when both

• PBS: a PVE cluster, LXC, a second-site sync, no Veeam budget, ransomware ACLs.
• Veeam: a Veeam console already covers vSphere/Hyper-V/Nutanix, you want one pane and application item restore, and you accept no LXC.
• Both: Veeam on Windows VMs with SQL/Exchange, PBS for the rest and for CTs. That is not a contradiction — it is two RPO targets.

More plug-in limits

Helpcenter (Considerations and Limitations), beyond LXC and templates:

• No IPv6, no MFA on the account used for PVE, no SSH Private Key credentials, no S3 as PVE storage, no OCSP on the certificate.
• A cluster is not added as an entity — each node separately. After PVE changes, sync with the backup server can take up to 15 minutes (or a rescan).
• VM permissions (users, groups, API tokens) are not in the backup. Restore of HA settings — no.
• Backup of VMs on LVM with Snapshots as Volume Chain (SAVC) enabled — yes. Workers and IR on LVM SAVC — no. Worker storage must be file-level with snapshots.

Instant Recovery and replication

IR to a PVE host has experimental support status in the helpcenter (separate article “Instant Recovery of Workloads to Proxmox VE”). The IR target needs QCOW2-capable storage; LVM SAVC as destination — no. IR from PVE backups also runs to vSphere / Hyper-V / AHV if those hosts are in the Veeam infrastructure. PVE replication is in the desktop console, does not keep the MAC for a replica in the same cluster, and BIOS UUIDs are rewritten per Veeam’s table.

PBS does not impersonate Veeam Instant Recovery: restore + start VM is a different RTO path. For LXC and second-site sync, PBS stays. Broader landing: ransomware-resistant backup.

Workers and file-level storage

Veeam helpcenter: the plug-in needs at least one file-level storage in PVE (directory, NFS, CIFS — not LVM/RBD alone). Workers also need file-level storage with snapshots. LVM with SAVC can be backed up, but a worker and Instant Recovery onto LVM SAVC — no. That is why a Ceph-RBD-only design still adds a file shelf for Veeam, or we stay on PBS.

The cap of 4 concurrent backups per storage is changed by Support, not a GUI slider. We do not “tune” workers beyond Veeam’s guidance. PBS has no workers in that sense: the client sits in PVE, the server accepts chunks. PBS scale is the datastore, the network and the job count, not a farm of worker VMs.