Docker

Buildx builders

Quick setup of an amd64 + arm64 builder

docker buildx create --bootstrap --driver docker-container --name bob-multi-arch --platform linux/amd64,linux/arm64 --use

The --use at the end selects it as your default builder, and can be omitted, and changed later with:

docker buildx use <builder name>

This works for most things, but for some things qemu emulators are needed.

Qemu emulators and binfmt

So to build multiarch on a given machine, buildx is used.

But sometimes it can fail with:

ERROR: failed to solve: process "/dev/.buildkit_qemu_emulator

In that case check the emulators by running:

docker run --privileged --rm tonistiigi/binfmt

Which should output something like:

{
  "supported": [
    "linux/amd64",
    "linux/arm64",
    "linux/riscv64",
    "linux/ppc64le",
    "linux/s390x",
    "linux/386",
    "linux/mips64le",
    "linux/mips64",
    "linux/arm/v7",
    "linux/arm/v6"
  ],
  "emulators": [
    "cli",
    "jar",
    "python3.10",
    "python3.11",
    "python3.12",
    "python3.13",
    "python3.8",
    "python3.9",
    "qemu-aarch64",
    "qemu-arm",
    "qemu-mips64",
    "qemu-mips64el",
    "qemu-ppc64le",
    "qemu-riscv64",
    "qemu-s390x"
  ]
}

If it outputs:

{
  "supported": [
    "linux/amd64",
    "linux/386"
  ],
  "emulators": null
}

Install emulators with:

docker run --privileged --rm tonistiigi/binfmt --install all

Or only the arm64 one:

docker run --privileged --rm tonistiigi/binfmt --install aarch64

Uninstall one:

docker run --privileged --rm tonistiigi/binfmt --uninstall qemu-aarch64

Uninstall all:

docker run --privileged --rm tonistiigi/binfmt --uninstall qemu-*

More info

Docker can build using an image in the registry as cache

Specifying external cache sources

Speedy builds with Docker Layer Caching

Docker-in-docker

Running docker with -v /var/run/docker.sock:/var/run/docker.sock is NOT docker-in-docker.

Docker runs directly on the host system, which has implications. (https://stackoverflow.com/a/55849875)

Real Docker-in-docker is running the special dind image, which runs a “real” docker inside the docker.

docker run --privileged -d docker:dind

Docker Compose:

• Releases

Network

• https://docs.docker.com/engine/reference/commandline/network_create/

• https://medium.com/@havloujian.joachim/advanced-docker-networking-outgoing-ip-921fc3090b09

• Container Network Model or CNM
• bridge vs macvlan
• macvlan vs ipvlan
• VLAN via macvlan
• macvlan and ipvlan

Host access to macvlan:

• Using Docker macvlan networks

In comments: I’ve been pulling my hair out trying to avoid macvlans, and use existing linux bridges.

docker network create --driver=bridge --ip-range=192.168.1.144/28 --subnet=192.168.1.0/24 --gateway=192.168.1.1 --attachable -o "com.docker.network.bridge.name=br0" br0

It was annoying because docker would clobber my br0 bridge and overwrite my local IP on br0 with whatever I set in gateway. My local machine is not the gateway for br0. In the end I went old school and tried starting docker with the args I used with “docker -d” previously eg:

dockerd --fixed-cidr=192.168.1.144/28 -b br0 --default-gateway=192.168.1.1

I noticed after inspecting the default bridge from that configuration that it was setting an aux address “DefaultGatewayIP”, so I took the command line options out, and created a new network like this:

docker network create --driver=bridge --ip-range=192.168.1.144/28 --subnet=192.168.1.0/24 --gateway=192.168.1.50 --attachable -o "com.docker.network.bridge.name=br0" --aux-address="DefaultGatewayIPv4=192.168.1.1" br0

Hey presto it works, my brctl defined bridge work harmoniously with my containers :-D

Advanced brdige config:

• Source

If you want your containers to join a pre-existing network, use the external option:

networks:
  default:
    external:
      name: my-pre-existing-network