[BUG] docker-compose - windows - Cannot connect to the Docker daemon at unix:///var/run/docker.sock

[doublefx]

Describe the bug
Not sure whether it is a missconfiguration or a bug

To reproduce
Steps to reproduce the behavior:

1. Create a docker-compose.yaml:

services:
  nginx-proxy:
    image: nginxproxy/nginx-proxy
    container_name: nginx-proxy
    ports:
      - "80:80"
    volumes:
      - /var/run/docker.sock:/tmp/docker.sock:ro
  autok3s:
    image: cnrancher/autok3s:v0.6.0
    init: true
    ports:
    - 8080
    volumes:
      - /var/run/docker.sock:/tmp/docker.sock:ro
      - C:\\Users\\doublefx\\.autok3s\\:/home/.autok3s/
    environment:
    - AUTOK3S_CONFIG=/home/.autok3s/
    - VIRTUAL_HOST=autok3s.vcap.me

2. run docker-compose -d up
3. create a k3d cluster
4. See an error: Cannot connect to the Docker daemon at unix:///var/run/docker.sock

OS: Windows 11 / Docker Desktop v4.13.1 (WSL2)
AutoK3s Version: v0.6.0

[Jason-ZW]

@doublefx Thanks for your feedback!

Try changing the volume mounts path /var/run/docker.sock:/var/run/docker.sock:ro like this:

services:
  nginx-proxy:
    image: nginxproxy/nginx-proxy
    container_name: nginx-proxy
    ports:
      - "80:80"
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
  autok3s:
    image: cnrancher/autok3s:v0.6.0
    init: true
    ports:
    - 8080
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - C:\\Users\\doublefx\\.autok3s\\:/home/.autok3s/
    environment:
    - AUTOK3S_CONFIG=/home/.autok3s/
    - VIRTUAL_HOST=autok3s.vcap.me

[doublefx]

I tried but it failed to start the nginx service.

Weirdly, when using /tmp/docker.sock and connecting to Container autok3s-autok3s-1, then running:

curl --unix-socket /tmp/docker.sock http://0.0.0.0/version

It works:

{"Platform":{"Name":"Docker Desktop 4.13.1 (90346)"},"Components":[{"Name":"Engine","Version":"20.10.20","Details":{"ApiVersion":"1.41","Arch":"amd64","BuildTime":"2022-10-18T18:18:35.000000000+00:00","Experimental":"false","GitCommit":"03df974","GoVersion":"go1.18.7","KernelVersion":"5.15.68.1-microsoft-standard-WSL2","MinAPIVersion":"1.12","Os":"linux"}},{"Name":"containerd","Version":"1.6.8","Details":{"GitCommit":"9cd3357b7fd7218e4aec3eae239db1f68a5a6ec6"}},{"Name":"runc","Version":"1.1.4","Details":{"GitCommit":"v1.1.4-0-g5fd4c4d"}},{"Name":"docker-init","Version":"0.19.0","Details":{"GitCommit":"de40ad0"}}],"Version":"20.10.20","ApiVersion":"1.41","MinAPIVersion":"1.12","GitCommit":"03df974","GoVersion":"go1.18.7","Os":"linux","Arch":"amd64","KernelVersion":"5.15.68.1-microsoft-standard-WSL2","BuildTime":"2022-10-18T18:18:35.000000000+00:00"}

So, I'm puzzling at figuring out why the create cluster cannot communicate with docker via sock.

[Jason-ZW]

This should be a bug in k3d provider. Only the default socket path is considered here, and the custom DOCKER_HOST env is not open to configure.
You can try to use the autok3s -d serve command to get it up and running.

[doublefx]

If you mean from directly from the CLI, see my other issue

[Jason-ZW]

Try this docker compose(which only running autok3s container with host port mapping), then access ui using http://127.0.0.1:<host mapping>:

services:
  autok3s:
    image: cnrancher/autok3s:v0.6.0
    init: true
    ports:
    - 8080
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - $HOME/.autok3s/:$HOME/.autok3s/
    environment:
    - AUTOK3S_CONFIG=$HOME/.autok3s/

[doublefx]

Yes, got it working with:

services:
  autok3s:
    image: cnrancher/autok3s:v0.6.0
    init: true
    ports:
    - 80:8080
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - $HOME/.autok3s/:/home/.autok3s/
    environment:
    - AUTOK3S_CONFIG=/home/.autok3s/

Do you know any other way to get a virtual host on the top?

[doublefx]

All right, a bit better:

services:
  nginx-proxy:
    image: nginxproxy/nginx-proxy
    container_name: nginx-proxy
    ports:
      - "80:80"
    volumes:
      - /var/run/docker.sock:/tmp/docker.sock:ro
  autok3s:
    image: cnrancher/autok3s:v0.6.0
    init: true
    ports:
    - 8080
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - $HOME/.autok3s/:/home/.autok3s/
    environment:
    - AUTOK3S_CONFIG=/home/.autok3s/
    - VIRTUAL_HOST=autok3s.vcap.me

Which allow to connect to autok3s.vcap.me and create a cluster but unfortunately the cluster does not appear in the UI.

But it still created:

And can connect to it from >_ Launch Kubectl

[Jason-ZW]

I will support enviroment DOCKER_HOST feature as soon as possible to support this scenario.

[doublefx]

Yes please as not having the cluster in the UI, I cannot do any action on it (Delete, Clone, ...)

@Jason-ZW
Actually after a bit, I refreshed and seen the cluster in the cluster list, so I:

1. Deleted it
2. Recreated the cluster
3. Managed to Launch Kubectl from the UI
4. Enabled explorer
5. connected to it.

[doublefx]

Also, I wonder if it has some connectivity consequences as none of the create pods succeeded, see FailedCreatePodSandBox in the above log, what do you think?

[Jason-ZW]

Can your environment access the Internet?
I try on my environment and the pods can be pulled down.

Unfortunately I think 0.6.0 is a failed version, it is recommended to use 0.5.2 to experience. Will release 0.6.1 ASAP.

[doublefx]

I can manually curl https://auth.docker.io/token?scope=repository%3Arancher%2Fpause%3Apull&service=registry.docker.io from inside the cluster

[Jason-ZW]

I will try your steps on windows tomorrow, then give your feedback message.

[Jason-ZW]

@doublefx Using Windows 11 22H2 can not reproduce your situation.

services:
  nginx-proxy:
    image: nginxproxy/nginx-proxy
    container_name: nginx-proxy
    ports:
      - "80:80"
    volumes:
      - /var/run/docker.sock:/tmp/docker.sock:ro
  autok3s:
    image: cnrancher/autok3s:v0.6.0
    init: true
    ports:
    - 8080
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - $HOME/.autok3s/:/home/.autok3s/
    environment:
    - AUTOK3S_CONFIG=/home/.autok3s/
    - VIRTUAL_HOST=autok3s.vcap.me

[doublefx]

Hi @Jason-ZW

I would love to see the same on my side but the server node does not seem to have any connectivity to the outside word:

E1115 18:10:12.787117       7 remote_runtime.go:116] "RunPodSandbox from runtime service failed" err="rpc error: code = Unknown desc = failed to get sandbox image \"rancher/pause:3.1\": failed to pull image \"rancher/pause:3.1\": failed to pull and unpack image \"docker.io/rancher/pause:3.1\": failed to resolve reference \"docker.io/rancher/pause:3.1\": failed to do request: Head \"https://registry-1.docker.io/v2/rancher/pause/manifests/3.1\": dial tcp: lookup registry-1.docker.io: Try again"

[Jason-ZW]

This could be a network issue in your environment.

[doublefx]

I will try today with the last RC