WIFIonICE vs Docker: Fixing DB (Deutsche Bahn) WIFI by moving Docker away from 172.18.x.x in /etc/docker/daemon.json

etc-docker-daemon.json

{
    "bip": "172.39.1.5/24",
    "fixed-cidr": "172.39.1.0/25",
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

Unfortunately, this did not work for me. Afterwards, docker did not run at all until I realised that /etc/docker/daemon.json must either

• not exist at all OR
• contain at least {} to make it valid json (cf. https://stackoverflow.com/questions/45335316/unable-to-configure-the-docker-daemon-with-file-etc-docker-daemon-json-eof)

@arne-cl Just to clarify: changing the bip and fixed-cidr subnet ranges didn't work? For the fix, the file is assumed to stay identical otherwise.

Dear @sunsided,
I deleted existing networks with docker network rm ... first, then saved your snippet verbatim in /etc/docker/daemon.json (I didn't have that file before).

Afterwards, docker wouldn't work, i.e.

docker run hello-world
docker: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?.
See 'docker run --help'.

Running sudo systemctl restart docker.service didn't help either. I also tried it without the nvidia part:

{
    "bip": "172.39.1.5/24",
    "fixed-cidr": "172.39.1.0/25"
}

but the result was the same.

it just worked for me, thanks!

Removing all networks first and leaving out the Nvidia config worked for me, thanks!

With the above solution, each time when connecting to WIFIonICE I was forced to remove networks created by docker compose again – which invalidated images and lead to running docker prune way too often…

This works for a couple of months now, without the need to prune networks each time (only once after initial setup):

/etc/docker/daemon.json

{
    "bip": "172.39.1.5/24",
    "fixed-cidr": "172.39.1.0/25",
    "default-address-pools":
    [{
        "base": "172.39.0.0/16",
        "size": 24
    }],
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    }
}

Doesn't work for me =(

The address for docker0 complies, but the br-<somethin> networks continue to use the 172.18.x.x range

Docker version 23.0.4, build f480fb1

Edit:

Ok, just got it working by running docker network prune. Networks created after that will use the new range

Nice! 🚀

I had to delete all already existing container and images. After i did that, it worked.