DinD with docker-compose

[augustobmoura]

After following the guide in https://hub.docker.com/_/docker/ I wrote this docker-compose.yaml that should setup the same as the shell commands (if I understood it correctly):

version: "3.8"

services:
  dind:
    image: docker:dind
    privileged: true
    environment:
      DOCKER_TLS_CERTDIR: /certs
    networks:
      - docker
    volumes:
      - docker-certs-ca:/certs
      - docker-certs-client:/certs/client

  docker-client:
    image: docker
    environment:
      DOCKER_TLS_CERTDIR: /certs
    # Just loop at infinitum, we will get to this container by `docker exec`ing
    command: [sh, -c, "while sleep 1; do :; done"]
    networks:
      - docker
    depends_on:
      - dind
    volumes:
      - docker-certs-client:/certs/client:ro

volumes:
  docker-certs-ca:
  docker-certs-client:

networks:
  docker:

After running docker-compose exec docker-client docker version I got the following error, and I cannot connect to the dind docker from the docker-client:

Client: Docker Engine - Community
 Version:           20.10.5
 API version:       1.41
 Go version:        go1.13.15
 Git commit:        55c4c88
 Built:             Tue Mar  2 20:14:11 2021
 OS/Arch:           linux/amd64
 Context:           default
 Experimental:      true
Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?

Is there anything I'm missing in my docker-compose.yml?

Running the commands from the README manually I get the correct result

[tianon]

You're missing docker-entrypoint.sh which will set DOCKER_HOST (and TLS-related variables) appropriately:

$ docker-compose exec docker-client docker-entrypoint.sh docker version

[augustobmoura]

Using docker-entrypoint shows another error:

Client: Docker Engine - Community
 Version:           20.10.5
 API version:       1.41
 Go version:        go1.13.15
 Git commit:        55c4c88
 Built:             Tue Mar  2 20:14:11 2021
 OS/Arch:           linux/amd64
 Context:           default
 Experimental:      true
error during connect: Get https://docker:2376/v1.24/version: dial tcp: lookup docker on 127.0.0.11:53: server misbehaving

[tianon]

Ohhh, the difference is that your DinD container is named dind instead of docker -- after renaming that, I was able to get it working. 👍

Alternatively, I reverted back to your verbatim file, added hostname: dind to the Docker-in-Docker container, and added DOCKER_HOST: tcp://dind:2376 to the client container, and that also works.

[augustobmoura]

Got it, so the problem is that I was not calling docker-entrypoint.sh which setups the env var DOCKER_HOST for a default tcp://docker:2376 url. Without the entrypoint I can just define the var to tcp://${docker_host}:${port} (and DOCKER_TLS_VERIFY DOCKER_CERT_PATH for tls support) and I'm good to go. Maybe we should put this on the README?

Thanks for the help!