VM corruption during PBS backups – still an open issue in 2025?

[Bob67]

Hello everyone,

we’re managing an on-premise Proxmox VE environment in an enterprise setting, running 35+ Windows Server VMs including Active Directory, SQL-based ERP, and critical file servers. We’re using Proxmox Backup Server (PBS) for regular backups and replication across two nodes.

Recently, we came across multiple threads and external reports (some still recent) describing VM corruption or boot issues (especially with Windows guests) after PBS backup jobs, particularly when using:

• Snapshot mode with QEMU guest agent (with fs-freeze enabled)
• High-IO workloads (SQL, AD, etc.)
• Backups interrupted during execution (hardware errors, crashes, or manual stop)

These incidents include:

• Filesystem corruption or unbootable Windows VMs
• Reboot loops after nightly backup jobs
• Inconsistent partitions or disk headers after backup

This is extremely critical in enterprise environments, where backup consistency is mandatory and VM corruption could cause serious downtime or data loss.

We’d like to ask the community and developers:

1. Are these issues still confirmed/valid in PBS 3.3+ and PVE 8.2/8.3?
2. Are there official workarounds or recommendations (e.g., disable fs-freeze, use suspend mode, run pre/post backup scripts)?
3. Is this considered a bug, usage risk, or unsupported behavior in specific configurations?
4. Would the Proxmox team consider making “safe mode” or “application-aware backups” more explicit in the UI for enterprise environments?

This is not meant as criticism, but rather as a request for clarity and best practice guidance, to help all users running PBS in business-critical environments make safe decisions.

Thanks in advance for your input and for the incredible work on the Proxmox stack!

Best regards,

Roberto M.

 

[_gabriel]

Main issue is when PBS is too slow,
like doing backup directly over WAN.
Or like multi backups to one HDD.

If you use incremental SQL backup system, set the VssOption option to the numerical value of 5 in Windows registry to not break the SQL chain backup.

I

Post in thread 'PBS breaking customer SQL backups. Backups without FS-Freeze?'

Jul 5, 2024

Ok, found it:

VSS-Options

Implementation, put the desired backup type number inside Regkey value
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\QEMU Guest Agent VSS
Provider\VssOption, so that the program can query the desired backup type.

VSS backup types:
number   type
1        VSS_BT_FULL
2        VSS_BT_INCREMENTAL
3        VSS_BT_DIFFERENTIAL
4        VSS_BT_LOG
5        VSS_BT_COPY

• ITT

• 

 

[Bob67]

Hi _gabriel,

well, we use 2x10Gbs Lacp interfaces on HPe DL380Gen10+ PVE(4) and HPe DL380Gen8 PBS(2) hardisk radi5 with dedicated 10Gbs switch.
I hope this can be ok.
Yes, we backup also VM Windows with SQL services on it, and ok, I set option numerical 5 for VssOpption.
(thanks)

I wait for a reply (I hope) from Proxmox developer/support.

best regards
Roberto M.

 

[fiona]

Hi,
you can use backup fleecing to mitigate these issues: https://pve.proxmox.com/pve-docs/chapter-vzdump.html#_vm_backup_fleecing