Proxmox VE host backup - best option for now

A

athompson3117

New Member
Nov 10, 2025
17
0
1
Brand new to ProxMox, but I've had 25+ years working on Linux machines as an Oracle Exadata (and other) DBA. I have my new PVE host set up, the network is configured how I want it, and I even have PBS running in a VM (it's a homelab, don't shoot me). Now I want a way to save all my hard work so that in the event of some kind of calamity (or I screw something up), I can restore to a working system without having to redo that past 48 hours.

To that end, there doesn't seem to be a good way to back up the host machine install, or at least a well-documented one. I found this:

https://pve.proxmox.com/wiki/Proxmox_Cluster_File_System_(pmxcfs)#_recovery

Seems straightforward, but based on other forum posts it also seems incomplete. Then there's this new entry:

D

Post in thread '[Script] Proxmox VE & PBS – System, Configuration and Structure and More Backup & Recovery'

athompson3117 said:
This sounds great. I just installed PVE yesterday as a replacement for my ESXi homelab, and very much would want something that can back up the host node. I'll keep an eye on this thread as feedback comes in.
Click to expand...
Hi! If you encounter any problems or bugs, please report them in the appropriate section of GitHub so that we can deal with them as effectively as possible.

But I'm still green when it comes to PVE. So I don't think I'm ready to be QA for an undertaking this complicated.

This forum post had a lot of good information, but was hard for me to follow (at least for now).

P

[SOLVED] Thread 'Official way to backup proxmox VE itself?'

Is there any official guide on how to backup proxmox VE itself?

My main concern is the amount of work that I have put in to make the current setup work, e.g. GPU passthrough etc.
If that breaks due to updates etc, I would like to have a way to quickly revert back to the last known working state.

Is Proxmox backup server able to do it? and is it the only way?
Or is there some way that is built-in to Proxmox VE itself?

So I thought I'd ask the community, based on my criteria, what would be the recommended "best for now" approach for me?

Thanks.
 
for me, the most reliable solution right now is Clonezilla. last week I restored everything with Clonezilla without any problems
 
gfngfn256 said:
See my post, on that thread you linked. It is still my current approach.
Click to expand...
That script looks like what I might be wanting. I noticed it hasn't been updated in a year. Does it work with the current version? Does it restore the full configuration (network, storage, backup targets, vm/docker configs, etc.)? If so, then this will probably do for PVE, and then I can use PBS for the the VMs (except PBS itself, which I'll just use direct backups for that).

I don't want to use a block-based clone as my VMs will be on the same drive(s) as PVE itself (a pair of 1TB SSDs in ZFS RAID 1). I suspect this will also preclude the Clonezilla approach which viktoria recommended. I've never used it but I'm guessing its a block-based backup. I used something similar for my ESXi server from which I'm migrating, but that was running off a simple thumb drive and everything else was stored on iSCSI LUNs on the NAS. With this new build I decided to move away from iSCSI and return to local storage now that SSDs are relatively cheap. I can still use the NAS for network shares to the VMs when I need bulk storage on them.
 
Backing up PVE in itself is not all that useful or even desirable, and should be downright avoided if a member of a cluster- cluster members are cattle not pets.

If you DO want to back up the host (lord knows why) any linux backup method would work- someone mentioned clonezilla.
 
  • Like
Reactions: UdoB
alexskysilk said:
Backing up PVE in itself is not all that useful or even desirable, and should be downright avoided if a member of a cluster- cluster members are cattle not pets.

If you DO want to back up the host (lord knows why) any linux backup method would work- someone mentioned clonezilla.
Click to expand...

Perhaps that's the case. My setup is not running as a cluster, and never will be run as a cluster. With that in mind, let's say something catastrophic happens to the host. What's the recovery method of choice? Build a new one from scratch and manually reconfigure it?

I guess what I had in mind was something similar to what you have with ESXi: you install the same version you were running onto new hardware (or reinstall on same hardware), and then restore the configuration backup. You're back up and running with all your storage, network, and configuration settings just as you had them. You just need to perform a restore of the VMs themselves to get them up and going (or not if they were on a storage target that is still in tact). I was honestly surprised Proxmox didn't have something like this.
 
Last edited:
athompson3117 said:
With that in mind, let's say something catastrophic happens
Click to expand...
Divide "catastrophic" into 2 categories. one is with your virtual payload and one is with your infrastructure. Consequently, its good practice to keep payload and infrastructure separated in terms of storage.

when something catastrophic occurs to your payload, restore it using PBS. you will find that PBS can be useful even if its not for catastrophic recovery.

when its your infrastructure, your first line of defense is to keep your boot device mirrored. If the system dies, you should be able to bring it back using your existing boot device(s.) To deal with a total boot device failure, you have choices:

1. reinstall and restore from pbs
2. reinstall and import the datastore. for that, all you need to do is keep copies of
/etc/pve/storage.cfg
/etc/network/interfaces
/etc/pve/lxc/*
/etc/pve/qemu-server/*
and copy them back onto your new server
3. keep a disk image of your boot device. clonezilla works for that. heaviest in terms of backup usage but most convenient to restore- just be mindful that you're grabbing an image of the boot device and not all drives in the system.

I can tell you that in 15 years of running PVE it was never an issue- even when I really did lose a boot device it took an extra 10 minutes to recreate vm.conf files for the disk images. and that was ONCE, on a non critical lab system I was ok using a single boot device that was faulty to begin with.
 
  • Like
Reactions: gfngfn256, news, athompson3117 and 1 other person
when something catastrophic occurs to your payload, restore it using PBS. you will find that PBS can be useful even if its not for catastrophic recovery.
Click to expand...
Agreed. I've set up PBS as a VM and will be using it to backup all the other VMs to a location on the NAS (there was a helpful YouTube video on how to do this for non-critical systems like mine using NFS)


when its your infrastructure, your first line of defense is to keep your boot device mirrored. If the system dies, you should be able to bring it back using your existing boot device(s.) To deal with a total boot device failure, you have choices:

1. reinstall and restore from pbs
2. reinstall and import the datastore. for that, all you need to do is keep copies of
/etc/pve/storage.cfg
/etc/network/interfaces
/etc/pve/lxc/*
/etc/pve/qemu-server/*
and copy them back onto your new server
3. keep a disk image of your boot device. clonezilla works for that. heaviest in terms of backup usage but most convenient to restore- just be mindful that you're grabbing an image of the boot device and not all drives in the system.
Click to expand...

#2 is probably what will serve me best given this is a homelab system and my configuration is unlikely to change very often at all. I should be able to set up a script to copy and zip these items to the NAS on a regular basis. It also sounds closest to have with my ESXi setup now. I might also add any sundry files located in /etc/pve/

So check my work on this. Using your terms...

Payload loss: Restore from PBS

Infrastructure loss:
1) Fix hardware (if necessary)
2) Reinstall PVE and any additional packages
3) Copy the records from #2 onto the new host
4) Restore PBS VM (backed up to NAS not using PBS)
5) Restore remaining VMs using PBS

Sound right?

I agree I would expect never to need to do any of this, but fortune favors the prepared.
 
athompson3117 said:
Does it work with the current version?
Click to expand...
Yes. See the adjustment(s) I made in that post.

athompson3117 said:
Does it restore the full configuration (network, storage, backup targets, vm/docker configs, etc.)?
Click to expand...
It backups all of that. I have never used it directly to restore. However you will literally have everything you need to reconstruct. I use it as the best documenting feature for your complete PVE setup & config available.

athompson3117 said:
I was honestly surprised Proxmox didn't have something like this.
Click to expand...
Well just read the forums - this rabbit-hole has been gone down numerous times.
In light of that, my personal guideline to follow is the following:
  • Do not to make any major system/OS changes to the Proxmox host itself.
  • Fully document ALL changes you make.
  • VMs & LXCs should be stored on a separate disk from the OS - where possible.
  • Have full & restorable backups of all VMs & LXCs available on an external medium.

alexskysilk said:
keep a disk image of your boot device. clonezilla works for that.
Click to expand...
I do that myself with dd & compress the image. (I've never had good experience with Clonezilla on a PVE host - I believe this will depend on your OS setup, zfs etc.). I make a full disk image (Host OS) before any major update (kernel etc). I have successfully restored from such an image.
Do note that THIS requires downtime on the server.
 
  • Like
Reactions: athompson3117
athompson3117 said:
It says in the post you "removed --one-file-system (twice) from script." Since is appears three times, which one did you leave in?
Click to expand...
I have just checked my (edited) script & I appear not to have it at all in my script. Possibly I removed it 3 times but forgot.

However I just checked the repo of DerDanilo/proxmox-stuff & it appears that the code now has it in 4 times! See here. So possibly he has made some changes. I can see some other changes - which appear to explain why he still has those --one-file-system in there. For context as to why I removed it - see here.
 
You must log in or register to reply here.
Top Bottom