This repository has been archived on 2026-07-30. You can view files and clone it. You cannot open issues or pull requests or push a commit.
Files
nixos/docs/pxe-boot.md
T
beatzaplentyandClaude Sonnet 5 ff021ca6f7
Check NixOS configurations / eval-hosts (push) Failing after 53m34s
Consolidate CI on codex-maintenance.sh; clean up dead scripts and docs
Both check-nixos.yml workflows (GitHub + Gitea) now call
scripts/codex-maintenance.sh instead of a hand-rolled eval-only loop,
closing a real gap: CI previously enforced none of the secret grep,
nixpkgs-fmt, or statix checks that codex-maintenance.sh already runs
locally — nothing was stopping that from regressing. One script now
backs both, instead of two copies that can drift from each other.

codex-maintenance.sh itself is extended to cover buildable surface
that wasn't validated anywhere before: packages.x86_64-linux.*, plus
config.system.build.tarball (lxc-* hosts) and
config.system.build.diskoImagesScript (proxmox-*, excluding the
installer's own proxmox-lxc target, which has no disko config).

Also:
- scripts/prepare-host-key.sh: dropped the redundant
  [path-to-nixos-repo] parameter — it always defaults to the repo the
  script itself lives in now, so a second argument never made sense
  after the nix-auto-installer migration.
- Removed prepare.sh (dead pre-disko manual parted/mkfs/mkswap
  partitioning, fully superseded) and
  scripts/create-linode-installer-disk.sh (incomplete draft for an
  abandoned dd-via-rescue-mode approach; Linode hosts already deploy
  fine through the normal auto-installer flow).
- docs/pxe-boot.md: fixed a stale `nixosConfigurations.pxe-boot` eval
  command (pre-refactor flat name, not a real flake attribute
  anymore) and added a cross-reference to docs/auto-installer.md.
- CLAUDE.md/README.md: full documentation pass reconciling this
  session's changes — modules/installer/, modules/pxe-boot/, the
  LXC/Proxmox image-building deployment paths, corrected the
  password-hash/SSH-key locations in the safety-rules section (both
  had drifted to reference files/paths that no longer exist), and
  added session-workflow guidance to prefer targeted host evals over
  full-repo sweeps for incremental changes (explicitly scoped to
  interactive sessions, not CI).

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01La55Nsss8jZ7ZuzUV9mfot
2026-07-20 07:02:23 +10:00

106 lines
3.2 KiB
Markdown

# pxe-boot
The `pxe-boot` host serves HTTP boot assets for iPXE clients — including a
self-staged copy of this flake's own auto-installer netboot image, see
`docs/auto-installer.md` for what that image actually is and does once
booted.
## Host Role
- Hostname: `pxe-boot`
- Web service: nginx on TCP port 80
- PXE root: `/srv/pxe`
- HTTP root for scripts and images: `/srv/pxe/http`
- TFTP root for first-stage bootloaders: `/srv/pxe/tftp`
- iPXE entry script: `/srv/pxe/http/boot.ipxe`
- Generated iPXE menu: `/srv/pxe/http/menu.ipxe`
- SystemRescue iPXE script: `/srv/pxe/http/systemrescue.ipxe`
- TFTP fallback script: `/srv/pxe/tftp/autoexec.ipxe`
- Boot binaries copied from the Nix `ipxe` package:
- `/srv/pxe/tftp/ipxe.efi`
- `/srv/pxe/tftp/undionly.kpxe`
## Directory Layout
The host creates these directories with systemd tmpfiles:
```text
/srv/pxe
/srv/pxe/http
/srv/pxe/http/images
/srv/pxe/http/nixos
/srv/pxe/http/systemrescue
/srv/pxe/http/ubuntu
/srv/pxe/http/rescue
/srv/pxe/tftp
```
Mount shared image storage under `/srv/pxe/http`, preferably
`/srv/pxe/http/images` unless a menu entry expects files in a specific
directory such as `/srv/pxe/http/nixos`.
The HTTP iPXE chain is:
```text
undionly.kpxe or ipxe.efi
-> autoexec.ipxe from the TFTP root, when iPXE requests it
-> http://192.168.2.247/boot.ipxe
-> http://192.168.2.247/menu.ipxe
```
The generated menu currently exposes entries for:
- NixOS installer
- SystemRescue environment
- iPXE shell
- Reboot
The NixOS installer entry chain-loads `/srv/pxe/http/nixos/netboot.ipxe`,
which is nixpkgs' own generated netboot iPXE script (correct `init=`/`initrd=`
kernel parameters included) rather than a hand-rolled boot line — that script
in turn expects its kernel/initrd siblings in the same directory. All three
files (`bzImage`, `initrd`, `netboot.ipxe`) are built from this flake's own
`modules/installer/iso.nix` netboot image (the same one `nix build .#pxe`
produces) and staged automatically by
`modules/pxe-boot/stage-installer-artifacts.nix` via `systemd.tmpfiles.rules`
— no manual operator step required.
The SystemRescue entry expects the source ISO at:
```text
/srv/pxe/http/images/systemrescue.iso
```
The `stage-systemrescue.service` oneshot extracts that ISO into:
```text
/srv/pxe/http/systemrescue
```
The rescue menu entry then chains `http://192.168.2.247/systemrescue.ipxe`,
which loads the SystemRescue kernel and initramfs from the extracted tree and
uses `archiso_http_srv` to fetch the squashfs payload over HTTP.
## Validation
Safe evaluation check:
```bash
nix eval .#nixosConfigurations.proxmox-pxe-boot.config.system.build.toplevel.drvPath --raw
```
After deployment by an operator, basic service checks are:
```bash
curl http://pxe-boot/boot.ipxe
curl http://pxe-boot/menu.ipxe
curl http://pxe-boot/systemrescue.ipxe
curl -I http://pxe-boot/systemrescue/sysresccd/boot/x86_64/vmlinuz
curl -I http://pxe-boot/systemrescue/sysresccd/boot/x86_64/sysresccd.img
```
During a successful BIOS chainload, TFTP should deliver `undionly.kpxe` once,
then nginx should log requests for `/boot.ipxe` and `/menu.ipxe`. Repeated TFTP
downloads of `undionly.kpxe` indicate the iPXE stage is still not reaching the
HTTP chain.