feat(qnap-lcd): Proxmox-correct front-panel LCD menu for esh-nas-pve; albok redeploy note

playbooks/qnap-lcd.yaml

@@ -0,0 +1,42 @@
+# Install the Proxmox-correct QNAP front-panel LCD menu on esh-nas-pve (Prime 2026-10-03: "it's running but not
+# correctly, can you fix?"). Idempotent. Originals kept beside: lcd-menu.py.upstream, /root/lcd-menu.service.orig.
+#   scripts/elway infra-ops@10.0.50.55 --playbook playbooks/qnap-lcd.yaml
+steps:
+  - name: keep the upstream menu script and unit
+    sudo: true
+    shell: |
+      set -e
+      cp -p /usr/local/bin/qnap_lcd/lcd-menu.py /usr/local/bin/qnap_lcd/lcd-menu.py.upstream
+      cp -p /etc/systemd/system/lcd-menu.service /root/lcd-menu.service.orig
+    creates: /usr/local/bin/qnap_lcd/lcd-menu.py.upstream
+  - name: menu script
+    sudo: true
+    upload:
+      src: services/qnap-lcd/lcd-menu.py
+      dest: /usr/local/bin/qnap_lcd/lcd-menu.py
+      mode: "0755"
+  - name: dry-run every screen before touching the panel
+    sudo: true
+    shell: python3 /usr/local/bin/qnap_lcd/lcd-menu.py --dry-run
+    changed_when: "false"
+  - name: service unit
+    sudo: true
+    upload:
+      src: services/qnap-lcd/lcd-menu.service
+      dest: /etc/systemd/system/lcd-menu.service
+      mode: "0644"
+  - name: re-enable (WantedBy moved to multi-user.target) and restart
+    sudo: true
+    shell: |
+      set -e
+      systemctl disable lcd-menu.service 2>/dev/null || true
+      systemctl daemon-reload
+      systemctl enable lcd-menu.service
+      systemctl restart lcd-menu.service
+verify:
+  - name: running, enabled, and no screen errors since the restart
+    sudo: true
+    shell: |
+      sleep 8
+      systemctl is-active --quiet lcd-menu.service && systemctl is-enabled --quiet lcd-menu.service && ! journalctl -u lcd-menu.service --since "-1min" --no-pager -q | grep -E "screen error|failed|Traceback"
+    changed_when: "false"

services/qnap-lcd/lcd-menu.py

@@ -0,0 +1,162 @@
+#!/usr/bin/env python3
+# QNAP front-panel LCD menu for esh-nas-pve (QNAP hardware running Proxmox VE, not QTS).
+#
+# Replaces the menu script of DigitalFunk/QnapLCD-Proxmox (checked out at /usr/local/bin/qnap_lcd, commit f2c9c46),
+# whose "Proxmox" fork still opened on the TrueNAS screen: menu[0] ran `cli -c 'system version'`, a TrueNAS-only
+# command, so the script died on its first screen and the panel stayed on "Initializing..." with dead buttons.
+# The same script also hid the management address: it skipped every bridge, and on Proxmox the IP lives on vmbr0.
+# Uses the repo's own `qnaplcd` driver unchanged (A125 display, 2x16, /dev/ttyS1 @ 1200 baud).
+# Source of truth: eshpfi services/qnap-lcd/ (deployed by playbooks/qnap-lcd.yaml). Test without the panel:
+#   python3 lcd-menu.py --dry-run      (prints every screen instead of writing to the LCD)
+import json
+import os
+import platform
+import subprocess
+import sys
+import threading
+import time
+
+sys.stdout.reconfigure(line_buffering=True)
+
+DISPLAY_TIMEOUT = 30   # seconds the backlight stays on after a button press
+REFRESH = 30           # seconds between redraws of the current screen
+PORT, PORT_SPEED = '/dev/ttyS1', 1200
+WIDTH = 16
+
+
+def shell(cmd):
+    return subprocess.check_output(cmd, shell=True, universal_newlines=True, stderr=subprocess.DEVNULL).strip()
+
+
+def fit(text):
+    return str(text)[:WIDTH]
+
+
+# ---------- screens: each returns two lines ----------
+
+def host_ip():
+    addrs = ipv4_addresses()
+    for name, ip in addrs:
+        if name == 'vmbr0':
+            return ip
+    return addrs[0][1] if addrs else 'no IPv4'
+
+
+def screen_host():
+    return [platform.node(), host_ip()]
+
+
+def screen_pve():
+    v = shell('pveversion')                       # pve-manager/9.2.21/4f6e0ac8 (running kernel: 7.0.14-20-pve)
+    ver = v.split('/')[1] if '/' in v else v
+    return [f'PVE {ver}', platform.release()]   # 'Proxmox VE 9.2.21' is 17 chars, one too many for the panel
+
+
+def screen_uptime():
+    secs = int(float(open('/proc/uptime').read().split()[0]))
+    d, rem = divmod(secs, 86400)
+    h, rem = divmod(rem, 3600)
+    m = rem // 60
+    up = f'{d}d {h:02d}:{m:02d}' if d else f'{h:02d}:{m:02d}'
+    load = os.getloadavg()
+    return [f'Up {up}', f'Ld {load[0]:.2f} {load[1]:.2f}']
+
+
+def ipv4_addresses():
+    out = []
+    for iface in json.loads(shell('ip -json address show')):
+        name = iface.get('ifname', '')
+        if name == 'lo' or name.startswith(('veth', 'tap', 'fwbr', 'fwpr', 'fwln')):
+            continue
+        for a in iface.get('addr_info', []):
+            if a.get('family') == 'inet':
+                out.append((name, a['local']))
+                break
+    return out
+
+
+def pools():
+    rows = shell('zpool list -H -o name,size,alloc,cap,health').splitlines()
+    return [r.split('\t') for r in rows if r.strip()]
+
+
+def build_menu():
+    menu = [screen_host, screen_pve, screen_uptime]
+    for name, ip in ipv4_addresses():
+        menu.append(lambda n=name, i=ip: [n, i])
+    for name, size, alloc, cap, health in pools():
+        menu.append(lambda n=name, s=size, a=alloc, c=cap, h=health: [f'{n} {h}', f'{a}/{s} {c}'])
+    return menu
+
+
+# ---------- driver glue ----------
+
+lcd = None
+lcd_timer = None
+menu = [screen_host]
+menu_item = 0
+lock = threading.Lock()
+
+
+def render(item):
+    try:
+        lines = menu[item]()
+    except Exception as e:                       # one broken screen must never kill the menu
+        lines = ['screen error', type(e).__name__]
+        print(f'screen {item} failed: {e!r}')
+    lines = [fit(l) for l in lines][:2]
+    if lcd is None:
+        print(f'[{item}] {lines[0]!r} | {lines[1] if len(lines) > 1 else ""!r}')
+        return
+    lcd.clear()
+    lcd.write(0, lines)
+
+
+def lcd_on():
+    global lcd_timer
+    lcd.backlight(True)
+    if lcd_timer:
+        lcd_timer.cancel()
+    lcd_timer = threading.Timer(DISPLAY_TIMEOUT, lambda: lcd.backlight(False))
+    lcd_timer.daemon = True
+    lcd_timer.start()
+
+
+def response_handler(command, data):
+    global menu_item
+    if command != 'Switch_Status':
+        return
+    with lock:
+        lcd_on()
+        if data == 0x01:      # up
+            menu_item = (menu_item - 1) % len(menu)
+        elif data == 0x02:    # down
+            menu_item = (menu_item + 1) % len(menu)
+        else:
+            return
+        render(menu_item)
+
+
+def main():
+    global lcd, menu, menu_item
+    if '--dry-run' in sys.argv:
+        menu = build_menu()
+        for i in range(len(menu)):
+            render(i)
+        return
+    sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+    import qnaplcd
+    lcd = qnaplcd.QnapLCD(PORT, PORT_SPEED, response_handler)
+    lcd.reset()
+    lcd.clear()
+    lcd_on()
+    while True:
+        with lock:
+            menu = build_menu()
+            menu_item %= len(menu)
+            render(menu_item)
+        time.sleep(REFRESH)
+
+
+if __name__ == '__main__':
+    main()

services/qnap-lcd/lcd-menu.service

@@ -0,0 +1,19 @@
+# QNAP front-panel LCD menu on esh-nas-pve. Source: eshpfi services/qnap-lcd/ (playbooks/qnap-lcd.yaml).
+# Replaces the repo's unit, which ran a shell wrapper with no restart: when the menu died (it always did on
+# Proxmox, see lcd-menu.py), nothing brought it back and the panel froze on "Initializing...".
+[Unit]
+Description=QNAP front-panel LCD menu
+After=network-online.target
+Wants=network-online.target
+
+[Service]
+Type=simple
+WorkingDirectory=/usr/local/bin/qnap_lcd
+ExecStartPre=/usr/bin/python3 /usr/local/bin/qnap_lcd/preinit.py
+ExecStart=/usr/bin/python3 /usr/local/bin/qnap_lcd/lcd-menu.py
+ExecStopPost=/usr/bin/python3 /usr/local/bin/qnap_lcd/shutdown.py
+Restart=on-failure
+RestartSec=10
+
+[Install]
+WantedBy=multi-user.target

stacks/albok-service/README.md

@@ -57,6 +57,11 @@ so `/srv/albok` is covered whole-VM. No file-level restic job for it.
 
 ## Deploy / redeploy
 
+⚠ **Do not restart the service in the middle of a Nemi walk.** The hourly walk on nh3-dev (`services/albok-nemi/`) fails with
+`Unreachable` and mails OnFailure; that happened 2026-10-03 16:11 at the 0.1.4 redeploy. It is harmless, since the next walk redoes the
+work and the service commits per batch, but avoidable. Check first, on nh3-dev as lkraven:
+`flock -n ~/.local/state/albok-nemi/run.lock true && echo idle || echo WALK RUNNING`
+
 ```bash
 # 1. host prep (idempotent)
 scripts/elway infra-ops@10.100.50.40 --playbook playbooks/albok-service-host.yaml