Docker images - issue

[blue-genie]

Hello,

I'm setting up a new raspberry with Industrial RS 485 and I'm not sure what I'm doing wrong:

If I use:

services:
  poolcontroller:
    group_add:
      - dialout
    devices:
      - /dev/ttyUSB0
    ports:
      - "4200:4200"
    image: ghcr.io/sam2kb/njspc

I get into an infinite loop of:

p1-poolcontroller-1  | Init state for Pool Controller
p1-poolcontroller-1  | [2/25/2026, 12:27:21 AM] info: Starting up SSDP server
p1-poolcontroller-1 exited with code 139

If I use:

services:
  poolcontroller:
    group_add:
      - dialout
    devices:
      - /dev/ttyUSB0
    ports:
      - "4200:4200"
    image: msmi/nodejs-poolcontroller

Everything starts up:

p1-poolcontroller-1  | Init state for Pool Controller
p1-poolcontroller-1  | /bin/sh: git: not found
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] warn: Unable to retrieve local git branch.  Error: Command failed: git rev-parse --abbrev-ref HEAD
p1-poolcontroller-1  | /bin/sh: git: not found
p1-poolcontroller-1  | 
p1-poolcontroller-1  | /bin/sh: git: not found
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] warn: Unable to retrieve local git commit.  Error: Command failed: git rev-parse HEAD
p1-poolcontroller-1  | /bin/sh: git: not found
p1-poolcontroller-1  | 
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] info: Starting up SSDP server
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] info: Checking njsPC versions...
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] info: Starting Pool System undefined
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] info: Listening for any installed OCPs
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] info: Auto-backup initialized Last Backup: 1970-01-01T00:00:00.000+0000
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] info: Server is now listening on 0.0.0.0:4200
p1-poolcontroller-1  | [2/25/2026, 12:30:51 AM] info: Serial port: /dev/ttyUSB0 request to open successful
p1-poolcontroller-1  | RESETTING DATA -- Data files backed up to ./logs directory.
p1-poolcontroller-1  | [2/25/2026, 12:30:58 AM] info: Initializing Nixie Control Panel for Nixie Single Body
...

As far as I know I should be using: ghcr.io/sam2kb/njspc

I also tried:

services:
   njspc:
      image: ghcr.io/sam2kb/njspc
      container_name: njspc
      restart: unless-stopped
      environment:
         - TZ=${TZ:-UTC}
         - NODE_ENV=production
         # Serial vs network connection options
         # - POOL_NET_CONNECT=true
         # - POOL_NET_HOST=raspberrypi
         # - POOL_NET_PORT=9801
         # Provide coordinates so sunrise/sunset (heliotrope) works immediately - change as needed
         - POOL_LATITUDE=28.5383
         - POOL_LONGITUDE=-81.3792
      ports:
         - "4200:4200"
      devices:
         - /dev/ttyUSB0:/dev/ttyUSB0
      # Persistence (create host directories/files first)
      volumes:
         - ./server-config.json:/app/config.json   # Persisted config file on host
         - njspc-data:/app/data                    # State & equipment snapshots
         - njspc-backups:/app/backups              # Backup archives
         - njspc-logs:/app/logs                    # Logs
         - njspc-bindings:/app/web/bindings/custom # Custom bindings
      # OPTIONAL: If you get permission errors accessing /dev/tty*, prefer adding the container user to the host dialout/uucp group;
      # only as a last resort temporarily uncomment the two lines below to run privileged/root (less secure).
      privileged: true
      user: "0:0"
volumes:
  njspc-data:
  njspc-backups:
  njspc-logs:
  njspc-bindings:

same outcome.

With the latest image it seem that if USB0 is detected, it fails to get past Starting up SSDP server - but there are no errors ...

Any idea what the issue might be?

[tagyoureit]

I can't reproduce this. I tried both images and they both started up for me. But I did update the Docker documentation. You want to be using ghcr.io/tagyoureit/njspc as the primary channel. That syncs every time I push up code.

[blue-genie]

Updated to use the recommended image, ghcr.io/tagyoureit/njspc, the issue persists. I'll see if I can troubleshoot it more.

Attaching to p1-poolcontroller-1
p1-poolcontroller-1  | Init state for Pool Controller
p1-poolcontroller-1  | [2/25/2026, 12:17:41 PM] info: Starting up SSDP server
p1-poolcontroller-1 exited with code 139

[blue-genie]

Stating with DEBUG=* node dist/app.js I get:

  socket.io:server initializing namespace / +0ms
  socket.io:server creating engine.io instance with opts {"allowEIO3":true,"cors":{"origin":true,"methods":["GET","POST"],"credentials":true},"cleanupEmptyChildNamespaces":false,"path":"/socket.io"} +2ms
  socket.io:server attaching client serving req handler +5ms
  express:router use '/socket.io-client' serveStatic +13ms
  express:router:layer new '/socket.io-client' +0ms
  express:router use '/' <anonymous> +1ms
  express:router:layer new '/' +0ms
[2/25/2026, 1:23:36 PM] info: Starting up SSDP server
  node-ssdp:server discovering all IPs from interface lo +0ms
  node-ssdp:server discovering all IPs from interface eth0 +1ms
  node-ssdp:server Will use interface eth0 +0ms
  serialport/stream opening path: /dev/ttyUSB0 +0ms
  serialport/bindings-cpp open +0ms
Segmentation fault

[blue-genie]

With help from grok it seems that there is an issue with alpine and rpi4 when it comes to serial bindings:

The container is Alpine Linux 3.23 (musl libc 1.2.5), running on your host's armv7l (32-bit Raspberry Pi kernel). Node v20.20.0 is fine, and serialport@11.0.1 / @serialport/bindings-cpp@11.0.3 is current—but the prebuilt native binary (bindings-cpp.node) for this exact combo (musl + armv7l/armhf) either doesn't exist or is broken/incompatible, causing the segfault exactly when the native code tries to open the serial port.

I was able to get in the image, as root, and do:

apk update
apk add --no-cache python3 make g++ gcc linux-headers git
cd /app
npm rebuild @serialport/bindings-cpp --build-from-source

manually ran node dist/app.js and it worked.

[blue-genie]

My final solution was to create a Dockerfile with:

FROM ghcr.io/tagyoureit/njspc AS build

# Work in the app directory (standard in most node images)
WORKDIR /app

FROM --platform=linux/arm/v7 node:20-bookworm-slim AS prod
..... same as in the Dockerfile from gituhub

[tagyoureit]

Oof. Good find. Glad it's working.