I migrated my first Proxmox system to a new server. It went very well. Basically you just back up all the containers and VMs, restore them, and Bob’s your uncle. Took under an hour, more because I did a test run installing to an external USB disk first. Really I’m migrating to the same server, I just wanted to switch from btrfs to zfs.
Proxmox backups include not only the guest OS image but also Proxmox’ own configuration. The restore is smart enough to adjust the config to, say, handle that the virtual disk image now lives in ZFS. The only place anything failed is that my VM had a virtual CDROM mounted; the restore still referenced the image via BTRFS. Then again the CDROM didn’t even exist in the new zfs server, I no longer needed it. No big deal: the error message was clear and I could easily edit the config to remove it.
The other hassle was restoring Proxmox itself. There’s no way to back up a Proxmox server, just its guests! I think they’re idea is you’re using a cluster with several servers and so if one fails you just add a new host and have it learn what to do. Still seems like they should have a backup option.
For a small home user it’s good discipline to do as very little on the Proxmox server itself as you can. I made a checklist of things to set up on the new one, here’s a quick list:
- update repositories and upgrade
- configure
/etc/fstabto mount my disks - configure backup disk as new storage type
- set up guest backup schedule in the GUI (again)
- add a system metrics server
apt install joe sudo avahi-daemon- install a user nelson for convenience but try not to use it
- install tailscale and remember
--accept-dns=false
I tried a couple of things to verify the new Proxmox was configured the same as the old without a lot of success. I was hoping that comparing /etc/pve would tell me more, that’s where Proxmox keeps its configs. But mostly it was stuff you’d expect to be different: crypto keys, details related to file system storage. Hard to find meaning in the noise. I was glad to find jobs.cfg though with a record of my backup configuration. (Note that /etc/pve is actually a pmxcfs FUSE filesystem for clusters; the actual truth lies elsewhere.)
Really pleased at how all this was. Virtualization is great! Next time will probably take half an hour at most and most of that time is waiting for disk images to copy around.
Nelson's log 