[BUG] v4.5.0 fails to start on a machine without IPv6 ("Address family not supported by protocol [core/socket.c line 82]")

[kripton]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

I updated from lscr.io/linuxserver/netbox:4.4.10 to lscr.io/linuxserver/netbox:4.5.0 and after the successful migrations, the uWSGI fails to start:

netbox  | *** Starting uWSGI 2.0.31 (64bit) on [Mon Jan 19 20:21:00 2026] ***
netbox  | compiled with version: 15.2.0 on 17 October 2025 09:22:48
netbox  | os: Linux-6.12.58-gentoo #1 SMP Wed Jan 14 10:56:03 UTC 2026
netbox  | nodename: 7192f1caea8b
netbox  | machine: x86_64
netbox  | clock source: unix
netbox  | pcre jit disabled
netbox  | detected number of CPU cores: 8
netbox  | current working directory: /app/netbox/netbox
netbox  | detected binary path: /usr/sbin/uwsgi
netbox  | your memory page size is 4096 bytes
netbox  | detected max file descriptor number: 1048576
netbox  | building mime-types dictionary from file /etc/mime.types...1390 entry found
netbox  | lock engine: pthread robust mutexes
netbox  | thunder lock: disabled (you can enable it with --thunder-lock)
netbox  | socket(): Address family not supported by protocol [core/socket.c line 82]
netbox  | [uWSGI] getting INI configuration from uwsgi.ini
netbox  | [uwsgi-static] added mapping for /static => static

My machine deliberately has IPv6 disabled on kernel level. I've had this with other docker images and the solution was to bind (inside the container) to 0.0.0.0 instead of localhost or ::1. However, with your image, I do not see any option to specify the host/IP to bind to.

Furthermore, I cannot go back to 4.4.10 since I am getting error messages now due to missing tables/columns, probably because of the migration.

Possibly related to #73 and #74

Expected Behavior

Netbox is up, running and usable

Steps To Reproduce

1. On a machine that has IPv6 disabled on kernel level
2. Run the image / compose project with default options

Environment

- OS: Gentoo Linux
- How docker service was installed: Docker compose file from 4 years ago, gradually updating the image until latest

CPU architecture

x86-64

Docker creation

services:
  netbox:
    # :latest (:4.5.x) fails to start due to missing IPv6 ("Address family not supported by protocol")
    image: lscr.io/linuxserver/netbox:4.5.0
    container_name: netbox
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=UTC
      - SUPERUSER_EMAIL=redacted
      - SUPERUSER_PASSWORD=redacted
      - ALLOWED_HOST=redacted
      - DB_NAME=netbox
      - DB_USER=netbox
      - DB_PASSWORD=redacted
      - DB_HOST=172.17.0.1
      - DB_PORT=5432
      - REDIS_HOST=172.17.0.1
      - REDIS_PORT=6379
      - REDIS_PASSWORD=redacted
      - REDIS_DB_TASK=1
      - REDIS_DB_CACHE=1
    volumes:
      - redacted:/config
    ports:
      - 127.0.0.1:8087:8000
    restart: unless-stopped

Container logs

see above

[github-actions]

Thanks for opening your first issue here! Be sure to follow the relevant issue templates, or risk having this issue marked as invalid.

[kripton]

Reference to the issue in "paperless-ngx" project where I had the same issue. However, they provide the configuration option PAPERLESS_BIND_ADDR to fix the issue. Settings its value to 0.0.0.0 makes paperless-ngx start: paperless-ngx/paperless-ngx#6546

[kripton]

Workaround found:
Assuming a volume mounting into /config exists and contains at least configuration.py:
Create a new volume mounting into /defaults and place the file https://github.com/linuxserver/docker-netbox/blob/9680c8d24a2704bbc0dcf1f95275c33574368f7d/root/defaults/uwsgi.ini there. In that file, change line http-socket = [::]:8000 to http-socket = 0.0.0.0:8000 and start the container.

UPDATE: And define that /defaults volume as read-only. Otherwise, the uwsgi.ini file will be MOVED and gone after one start of the container. Making the volume ro makes sure the file stays there and surprisingly does not block the container startup. I expected the mv command to fail but it doesn't complain.