I am unable to replicate my workflow where I would create a configuration.nix file of the virtual machine I want to create by using the following command.
It instead tries to apparently run the iPXE, which looks quite unexpected thing and as expectedly it fails and produces the…
> [ 104.912947] Rebooting in 1 seconds..
> [2026-02-05T17:52:44Z INFO virtiofsd] Client disconnected, shutting down
> [2026-02-05T17:52:44Z INFO virtiofsd] Client disconnected, shutting down
> Virtual machine didn't produce an exit code.
Which previously it just built the image and it didn’t produce anything. Looking in the /nix/store I don’t find anything to look into.
Is there a better way to have declarative NixOS builds in Proxmox or am I forced to keep the deprecated repository in use manually by the usual building it?
Also it looks like there is nothing to document about the disk-size variable as well when it was possible before.
error:
… while calling the 'attrNames' builtin
at «string»:6:3:
5| in
6| builtins.attrNames set.config.system.build.images
| ^
7|
… while evaluating the attribute 'config.system.build.images'
at /nix/var/nix/profiles/per-user/root/channels/nixos/nixos/default.nix:19:23:
18| {
19| inherit (eval) pkgs config options;
| ^
20|
(stack trace truncated; use '--show-trace' to show the full trace)
error: The option `virtualisation.proxmox' does not exist. Definition values:
- In `/home/nixosbm/manual_builds/forgejo.nix':
{
qemuConf = {
bios = "ovmf";
};
}
Which is weird because I would have expected this to be 1:1 nixos-generate package that was deprecated even if it was merged upstream.
Even though I can see this existing in the upstream kinda, but I don’t know how to exactly import it.
So in the meanwhile I though to well go to SeaBIOS then, but when the following settings…
# Bootloader - GRUB for Legacy BIOS
boot.loader.systemd-boot.enable = lib.mkForce false;
boot.loader.grub = {
enable = true;
device = "/dev/vda"; # Install GRUB to the disk
efiSupport = false; # Disable UEFI
};
boot.initrd.availableKernelModules = ["virtio_pci" "virtio_scsi" "ahci" "sd_mod"];
# Force Disk Size to 96GB to avoid Proxmox issues with resizing
virtualisation.diskSize = 96 * 954; # 96GB in MiB
It would fail with the following
ERROR: cptofs failed. diskSize might be too small for closure.
error: builder for '/nix/store/g9z0sgr88s0q24bm84286lyai6d03ngw-proxmox-nixos-25.11.5198.e576e3c9cf9b.drv' failed with exit code 1;
last 25 log lines:
> [ 35.066259] Mem-Info:
> [ 35.066263] active_anon:0 inactive_anon:0 isolated_anon:0
> [ 35.066263] active_file:7923 inactive_file:7924 isolated_file:0
> [ 35.066263] unevictable:0 dirty:47 writeback:0
> [ 35.066263] slab_reclaimable:5086 slab_unreclaimable:2842
> [ 35.066263] mapped:0 shmem:0 pagetables:0
> [ 35.066263] sec_pagetables:0 bounce:0
> [ 35.066263] kernel_misc_reclaimable:0
> [ 35.066263] free:1315 free_pcp:0 free_cma:0
> [ 35.066277] Node 0 active_anon:0kB inactive_anon:0kB active_file:31692kB inactive_file:31696kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:0kB dirty:188kB writeback:0kB shmem:0kB writeback_tmp:0kB kernel_stack:116kB pagetables:0kB sec_pagetables:0kB all_unreclaimable? yes
> [ 35.066286] Normal free:5260kB boost:4096kB min:5364kB low:5680kB high:5996kB reserved_highatomic:0KB active_anon:0kB inactive_anon:0kB active_file:31692kB inactive_file:31696kB unevictable:0kB writepending:188kB present:102400kB managed:100768kB mlocked:0kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
> [ 35.066314] lowmem_reserve[]: 0 0
> [ 35.066320] Normal: 19*4kB (U) 280*8kB (UE) 126*16kB (UE) 23*32kB (UE) 3*64kB (U) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 5260kB
> [ 35.066337] 15847 total pagecache pages
> [ 35.066342] 25600 pages RAM
> [ 35.066346] 0 pages HighMem/MovableOnly
> [ 35.066349] 408 pages reserved
> [ 35.066352] Tasks state (memory values in pages):
> [ 35.066357] [ pid ] uid tgid total_vm rss pgtables_bytes swapents oom_score_adj name
> [ 35.066363] Out of memory and no killable processes...
> [ 35.066390] Kernel panic - not syncing: System is deadlocked on memory
> [ 35.066397] ---[ end Kernel panic - not syncing: System is deadlocked on memory ]---
> cptofs: lib/posix-host.c:448: panic: Assertion `0' failed.
> /nix/store/89c2y24wj6z3q7ab153wc68qjgjbkn14-vm-run: line 132: 1379 Aborted (core dumped) cptofs -p -P 1 -t ext4 -i $diskImage $root/* /
> ERROR: cptofs failed. diskSize might be too small for closure.
Which is super weird that 96GB would not be enough, but strangest thing even I had put the config there, it would yell at me for this
trace: evaluation warning: Obsolete option `proxmox.qemuConf.diskSize' is used. It was renamed to `virtualisation.diskSize'.
Which means that something is very much up with the proxmox image building with nixos-rebuild build-image --image-variant proxmox
Apologies, when running the full-trace of the old build machine, it indeed does run iPXE, but only if I had run with full-trace
on my part at myself.
I guess TIL how it actually guarantees that it runs as it is supposed to.
Also another big derp and future warning for you all
virtualisation.diskSize = 98304; # 96GB in MiB
This will create an .vmz.zst file that is 103GB initially, so if you do not have this amount of diskspace…
You will be in a world of hurt, as it assumes you have that amount of diskspace. Which actually makes a lot of sense, as you it creates a backup file with everything included, including empty space.
Completely my bad on my part.
I will leave it open of what to do when people want to still use OVMF instead of SeaBIOS but unable to due to deprecation.
on my part at myself.