This article describes Proxmox VE configurations and behaviors that are either not supported or may result in unexpected backup and restore outcomes.
Use this as a reference when planning, troubleshooting, or validating Proxmox backup and restore operations.
Unsupported Configurations
These conditions should be treated as misconfigurations. They can lead to backup/restore failures, incorrect disk handling, or unexpected errors.
Pool level path
Linux block device paths such as /dev/data/test-disk-lvm-1 are not supported. Only Storage layer syntax (storage: vm-<VMID>-<Disk Name>) are supported.
Behavioral Limitations
These behaviors are by design and should be considered when planning capacity and operational procedures.
Disk size alignment during restore
ZFS-based restores do not support disk sizes with fractional MiB values. During restore, disk sizes are automatically rounded to the nearest whole MiB.
For LVM-Thin storage, restored disk sizes are aligned to the nearest 4 MiB boundary. The restored disk may be slightly larger than the original. This is expected behavior and is required for alignment on LVM-Thin volumes.
Backup proxy node migration
Migrating the backup proxy to a different Proxmox node is not supported. This can disrupt ongoing backup and restore operations, and can result in failed backups or restores due to loss of node-specific context.
Restore VM cleanup
Restore VM is not cleaned after proxy is restarted. You must clean it manually.
Partial restoration of VM properties
Some VM-level property values are not restored and revert to default values after restore. These properties are:
Start at boot
ACPI support
KVM hardware virtualization
Freeze CPU at startup
Protection flag
SPICE enhancements
VM state storage
AMD SEV
Intel TDX
Unsupported devices
Backup of CD/DVD device, USB device, PCI device, serial port, CloudInit drive, Audio device, VirtIOn RNG, VirtIOfs is not supported.
Backup Limitation
An overmapped state occurs when there is a mismatch between the virtual size assigned to an LVM thin volume and its actual metadata/allocation mapping table.
This condition occurs during disk shrinking operations. If a virtual disk's size is reduced within the hypervisor or volume manager, but the underlying LVM thin pool metadata is not properly synchronized or updated, the allocation table remains larger than the new disk boundaries. Consequently, the backup process may miscalculate block allocations, causing the entire backup job to abort. During volume scans (lvs) or backup initialization, the system will output a warning similar to the following:
WARNING: Thin volume pve/vm-140-disk-1 maps 1.50 GiB while the size is only 1.00 GiB.