ansible: edge-host role — bare-metal Arch N100 replaces k8s-edge VM #95

2026-07-02T10:04:01Z

k.renner commented

2026-07-02 10:04:01 +00:00

Prepares the dedicated Arch host (edge-1.srv.renner.uno) that replaces the k8s-edge Talos VM. Covers everything still pinned there plus the new CAN service:

zwave-js-ui + sml2mqtt via docker compose — configs ported 1:1 from the k8s manifests (same images, same by-id device paths, sml2mqtt config identical incl. Wh/entity-id continuity; --device instead of privileged hostPath)
open3e natively: can0 via systemd-networkd ([CAN] BitRate=250000, auto-restart on bus-off), venv + hardened-ish systemd unit bound to the CAN device, read-only 21-datapoint set, dedicated open3e broker user (PR #94)
inventory group edge_host, secrets via group_vars/edge_host/vault.yml (kubectl one-liners in the README), base-install + per-service cutover checklist in roles/edge-host/README.md
vm_153 tf comment updated to document the deprecation path

Not in scope (follow-up after cutover): removing the flux kustomizations, PVC, vm_153 and the Talos node patch.

Context: stock Talos ships CONFIG_CAN unset (siderolabs/talos#7686) — SocketCAN can't work in any pod without a custom kernel per release, hence bare metal.

🤖 Generated with Claude Code

Prepares the dedicated Arch host (`edge-1.srv.renner.uno`) that replaces the k8s-edge Talos VM. Covers everything still pinned there plus the new CAN service: - **zwave-js-ui** + **sml2mqtt** via docker compose — configs ported 1:1 from the k8s manifests (same images, same by-id device paths, sml2mqtt config identical incl. Wh/entity-id continuity; `--device` instead of privileged hostPath) - **open3e** natively: `can0` via systemd-networkd (`[CAN] BitRate=250000`, auto-restart on bus-off), venv + hardened-ish systemd unit bound to the CAN device, read-only 21-datapoint set, dedicated `open3e` broker user (PR #94) - inventory group `edge_host`, secrets via `group_vars/edge_host/vault.yml` (kubectl one-liners in the README), base-install + per-service cutover checklist in `roles/edge-host/README.md` - vm_153 tf comment updated to document the deprecation path Not in scope (follow-up after cutover): removing the flux kustomizations, PVC, vm_153 and the Talos node patch. Context: stock Talos ships `CONFIG_CAN` unset (siderolabs/talos#7686) — SocketCAN can't work in any pod without a custom kernel per release, hence bare metal. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

k.renner added 1 commit 2026-07-02 10:04:01 +00:00

ansible: add edge-host role — bare-metal Arch N100 replaces k8s-edge VM

ci/woodpecker/push/deploy-infrastructure Pipeline was successful

Details

42a5cdc275

Replaces the k8s-edge Talos VM (vm_153) with a dedicated Arch host
running the hardware-bound services: zwave-js-ui and sml2mqtt via
docker compose (configs ported 1:1 from the k8s manifests), plus the
new open3e CAN bridge (SocketCAN via systemd-networkd, venv + systemd
unit, read-only datapoint set).

Driver for leaving k8s here: stock Talos ships CONFIG_CAN unset
(siderolabs/talos#7686), so the Vitocal/VX3 CAN adapter cannot work in
any pod without a custom kernel per release.

Base-install procedure, secret bootstrap (vault.yml from cluster
secrets) and per-service cutover checklist in roles/edge-host/README.md.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 10:15:02 +00:00

images: add edge-host installer ISO ee98c88ef4

Releng-based archiso overlay: `edge-install` in the live env wipes the
N100 NVMe and installs the ansible-ready base system (systemd-boot,
networkd DHCP, sshd key-only, docker). Service config stays in the
edge-host ansible role; the ISO only produces the base.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 10:34:39 +00:00

images: build edge-host ISO in an archlinux container dbe8efdc62

mkarchiso against the host repos fails on Manjaro (no archinstall/
reflector, linux is a virtual package) — releng needs real Arch repos.
Split the profile assembly into build-inner.sh and run it in
archlinux:latest (--privileged for pacstrap's bind mounts); verified
build on the Manjaro workstation. gitignore the ISO output dir.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 10:38:28 +00:00

edge-host: open3e must publish JSON payloads (-j) 5586f4bfee

Without -j open3e flattens complex DIDs into per-key subtopics; the HA
package parses value_json on the base topic and receives nothing.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 10:43:07 +00:00

edge-host-iso: headless install support + internal-disk autodetect 81dc3d0a85

Target changed from N100 to a Wyse 3040 (2G RAM / 8G eMMC): default
disk is now autodetected (nvme, else mmcblk0, else first non-USB disk)
and the live environment carries the homelab SSH keys, so the install
runs over SSH without a monitor (firmware boot-order/Secure-Boot still
needs one console visit if not already set).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 10:55:41 +00:00

edge-host: extend open3e datapoint set to 39 DIDs 4576c2d13d

Adds PV energy counters (1802, Energy dashboard solar slot), battery
power/energy/temperature and dynamic charge limits, heating/DHW energy
split (electric in, thermal out — live COP), compressor modulation and
starts/hours, refrigerant inlet/outlet temps, buffer temp, DHW + valve
states. Selected from the full DID scans (539+189 DPs).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 11:36:34 +00:00

edge-host-iso: fix eMMC detection on the Wyse 3040 ed80d83a73

The live env detects the eMMC card (sdhci_acpi) but does not autoload
mmc_block, so /dev/mmcblk0 never appears and edge-install finds no
target. Autoload via modules-load.d, belt-and-braces modprobe in
edge-install, check mmcblk0-2, and fix the lsblk fallback (empty TRAN
column shifted the awk fields).

Diagnosed live over SSH on the actual 3040 (10.2.50.109).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 11:37:37 +00:00

edge-host: poll the VCMU (0x68c) refrigerant-circuit ECU — 53 DIDs total ecadc983be

Adds suction/discharge pressure, superheat, EEV positions, fan speeds,
compressor rps, four-way valve + circuit status, evaporator/economizer/
oil temps and defrost energy target, discovered via depictSystem ECU
census + full VCMU DID scan.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 11:48:46 +00:00

edge-host-iso: pin mmc modules into the installed initramfs 244929fff2

mkinitcpio's autodetect missed the eMMC stack in the chroot; the
installed system hung unbootable with root on mmcblk0p2. Pin
sdhci_acpi/sdhci_pci/mmc_block in MODULES and regenerate during install.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 12:07:44 +00:00

edge-host: edge-1 lives on the IoT VLAN — fix inventory FQDN f8d6025048

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 12:11:23 +00:00

edge-host: don't abort the play on not-yet-moved USB devices 89da3ef0d4

The cutover order moves devices one at a time; a compose service whose
stick is still on the old VM must not block open3e installation.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 12:13:18 +00:00

edge-host: pip module needs python-packaging on the target 46152c0bdf

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 12:17:17 +00:00

edge-host: force can0 up via ActivationPolicy=always-up 9e03cfa213

networkd applies [CAN] bitrate but does not raise the link by itself;
also make the handler reconfigure after reload (reload alone leaves
existing links untouched).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner added 1 commit 2026-07-02 12:23:52 +00:00

edge-host: open3e dir owned by service user (udsoncan log file) 4a49470cd2

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

k.renner merged commit dd63f372df into master

2026-07-02 12:27:50 +00:00

k.renner referenced this issue from a commit

2026-07-02 12:27:51 +00:00

Merge pull request 'ansible: edge-host role — bare-metal Arch N100 replaces k8s-edge VM' (#95) from edge-host-arch into master

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: konairius/homelab#95