Migration of PVE Single Host Filesystem with the help of PBS
This website and the other services run on top of a single node, single disk PVE (Proxmox Virtualisation Engine) server. At least for now. A cluster is already anticipated. When creating the hypervisor host, I made an error and created it with an ext4 lvm filesystem instead of a zfs. I wanted to change it to a zfs host which required a reinstallation of PVE.
Luckily, I already have a backup which made things easier.
Setup
I have:
- Single node, single disk PVE hypervisor
- 9 LXC on the PVE (1-2 GB disk per LXC)
- 1 VM on the PVE (8 GB disk)
- Single node, single disk PBS (Proxmox Backup Server) working and creating usable backups
Plan
My plan was to
- backup all vm and lxcs on PVE to PBS for having the best data set
- shutting down all guests to not alter any data afterwards
- verify backups on PBS
- Backup PVE config (e.g. screenshots of network settings, etc.)
- Reinstall PVE
- Reconfigure PVE
- Restore Guests
Implementation
The implementation was straight forward. I already had a working backup concept, so I just had to "Backup now" all guests. Afterwards, I turned off all guests and verified the latest backups on the PBS. This took me around five minutes overall. Then I manually screenshoted relevant configuration (network, backup storage, licensing, host, dns, etc.) of the PVE (~ two minutes). Reinstallation of the PVE took place via manually booting from a USB drive and install the system (classical installation). I was carefully picking the correct configuration now (~ five minutes), and then reconfiguring the clean PVE host with the settings from the previously taken screenshots (~ 10 minutes), including networks and backup storage (PBS). This enabled me to restore my guests again. How to restore guests, that are not available on the PVE yet? In the GUI via the backup storage or in the shell via commands.
LXC:
pct restore --storage <STORAGE> <ID> <backup-path>VM:
qmrestore --storage <STORAGE> <ID> <backup-path>
To restore the selected LXC from the image, the require command would be:
pct restore --storage local-zfs 108 primary-backup:backup/ct/108/2025-07-09T03:11:10Z
The PVE is supporting this process very well, as the tab lists existing, available backups. The restore was done in aroung five minutes and I was up and running again.
Summary
Restoring guests to PVE with PBS is surprisingly easy and user friendly. It is increadibly fast (the overall operation and therefore downtime was around an hour). Also, changing the filesystem does not have any visible impacts on guests behaviour. No problems occured but the wanted zfs features (compression, deduplication, etc.) are present.
Last update: 2025-07-09