I got it working - backtracked to the original image i was using 1 hr ago - gitea/act_runner:0.2.9 runs ok with docker:26.1.4-dind
So I used the config.yaml file you specify in your exampleā¦ and then use the ubuntu-22.04 label to run my loadbuild scripts. My scripts install docker, try to use buildx and the installation went fineā¦ but then buildx erroredā¦
ERROR: failed to initialize builder tender_blackwell (tender_blackwell0): Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
I added the āoptionsā section Iād originally missed to my original configmap and still see same type of errorā¦
ERROR: failed to initialize builder silly_dewdney (silly_dewdney0):
Cannot connect to the Docker daemon at unix:///var/run/docker.sock.
Is the docker daemon running?
The runner pod has the DOCKER_HOST set to tcp addressā¦
I use an image self-hosted in Gitea from here:
gitea/runner-images: Official Gitea Actions runner images - runner-images - Gitea: Git with a cup of tea, ubuntu-act:22.04
The Gitea action in my example is unchanged,
did you try this ?
I havenāt tried the whole example as-isā¦ I tried the configmap.
I believe Iām also use a self-hosted gitea act-runner image for the runner.
I then used gitea runner ubuntu22.04 you have in your labels, and I also tried my home-grown ubuntu 22.04 image. Both ubuntu versions have the same socket error.
I will try the gitea action next.
1 Like
I understand
but this label,
is not representing the stock ubuntu 22.04 image.
and in the action you can overide that, with:
jobs:
build-image:
name: build-image
runs-on: ubuntu-latest
container:
# image: ghcr.io/catthehacker/ubuntu:act-latest
# image: gitea/runner-images:ubuntu-latest
image: gitea.example.com/mariusrugan/ubuntu-act:22.04
The label just allows you to specify runs-on:
name: Build Service
on:
push:
branches: [main]
jobs:
pre-build:
runs-on: ubuntu-latest
outputs:
Yeah, I did change the runs-on to a gitea runner when I change the configmap file to the example you postedā¦
- "ubuntu-latest:docker://gitea/runner-images:ubuntu-latest"
- "ubuntu-22.04:docker://gitea/runner-images:ubuntu-22.04"
- "ubuntu-20.04:docker://gitea/runner-images:ubuntu-20.04"
I got the same socket error. I will double-check.
i understand.
but i think youāre not touching the image at all,
do this in your action:
image: gitea/runner-images:ubuntu-22.04
Arh, you meanā¦ using a container imageā¦ no, I am not. 
From your exampleā¦
container:
# image: ghcr.io/catthehacker/ubuntu:act-latest
# image: gitea/runner-images:ubuntu-latest
image: gitea.example.com/xyz/ubuntu-act:22.04
credentials:
username: ${{ gitea.actor }}
password: ${{ secrets.GA_TOKEN }}
1 Like
And also the stateful set, switch it to gitea/act_runner:0.2.9 because youāre using :latest and is not working.
The reason is the latest Dockerfile is changed, and tini doesnāt start
here is the reason: act_runner/Dockerfile at main - act_runner - Gitea: Git with a cup of tea
in the stateful set, the #1 container is looking for docker to start, and with :latest this contraption is not working, will fix it and post a message.
The idea was from here:
The fix is very simple:
replace the current command with:
command: ["sh", "-c", "while ! nc -z localhost 2376 </dev/null; do echo 'waiting for docker daemon...'; sleep 5; done; /sbin/tini -- run.sh"]
so image: gitea/act_runner:nightly runs with
Also you can upgrade Docker DinD image to: docker:27.3.1-dind
1 Like
Thanks for the info.
So, I have this workflowā¦ and Iāve tried with setup.docker.sh installing docker and not installing docker and it fails as soon as I call docker buildx create with the same message about socket.
jobs:
build:
runs-on: ubuntu-22.04
container:
image: gitea/runner-images:ubuntu-latest
steps:
- uses: actions/checkout@v4
- name: Setup Toolchain (Docker Buildx)
run: .gitea/workflows/helpers/scripts/setup.docker.sh
- name: Build and Push Contest Server Docs
env:
REGISTRY_LOADBUILDER_PASSWORD: ${{ secrets.REGISTRY_LOADBUILDER_PASSWORD }}
SSH_PRIVATE_KEY: ${{ secrets.SSH_PRIVATE_KEY }}
BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
API_VERSION: "dev"
RELEASE_NAME: "dev"
run: |
cd docs/sphinx/top-level/
./build-and-push.sh
It is launched now on a gitea runner imageā¦
act-runner-5587748d6d-vzdhs(version:v0.2.9) received task 4845 of job 4534, be triggered by event: push
workflow prepared
evaluating expression 'success()'
expression 'success()' evaluated to 'true'
š Start image=gitea/runner-images:ubuntu-latest
š³ docker pull image=gitea/runner-images:ubuntu-latest platform= username= forcePull=false
š³ docker pull gitea/runner-images:ubuntu-latest
...
What i recommend is always have a sleep in there that pauses the flow and hop in the container and debug it.
so at the run: | level
# while [ 1 -eq 1 ];do sleep 3600;done;
The āsetup.docker.shā runs just this nowā¦
echo "Creating buildx using $DOCKER_HOST"
docker buildx create
which givesā¦
Creating buildx using
ERROR: failed to initialize builder blissful_lumiere (blissful_lumiere0): Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?
Yeah, will do - I was doing that earlier with the runner and dind containers.
Thanks for all the suggestions and troubleshootingā¦ I figure I still must have some misconfiguration and will go over this more carefully.
1 Like
iirc
there is more to buildx to set it up correctly.
use the action, iād say
# https://github.com/docker/setup-buildx-action
# https://docs.docker.com/engine/reference/commandline/cli/#environment-variables
- name: Setup Docker Buildx
id: buildx
uses: docker/setup-buildx-action@v3
with:
# buildkitd-flags: --debug
endpoint: builders
env:
DOCKER_HOST: "tcp://docker:2376/"
DOCKER_TLS_CERTDIR: "/certs"
DOCKER_TLS_VERIFY: 1
DOCKER_CERT_PATH: "/certs/client"
Also this endpoint setup
1 Like
no problem, glad to help
āThe onionā is very complex, with many layers, i hate it too.
Github actions running nodejs this that context, endpoint, docker TCP address ā¦ Everything is worth writing a book 
I think youāre really close,
let me know if any help is needed here.
EDIT:
if you want to do your own thing, build on top of the image from Gitea, setup everything there, and just run one step and a script. Itās perfectly sane and fine.
Job succeeded
Ok, ok, okā¦ it works now! So, I think my problem was I tried to retrofit my original workflow with probably ~70% of your example and ignored some critical onion layers.
Most notably, the DOCKER_* ENVs and tried to home-bake the docker buildx setup instead of using the āSetup Docker Buildxā action. I had got quite far on a non-k8s environment but clearly had assumed a lot of magic.
I think there were ~2-3 onion layers I wasnāt even thinking about. I had just assumed a gitea runner image would be sufficient. I am not 100% sure why this container is neededā¦ I assume some magic with nodejs and tcp:// docker host?
But this seemed to be the major thing I was missing for anyone else who has made it this far in the threadā¦
container:
image: gitea/runner-images:ubuntu-latest
Thanks again @mariusrugan - Iām not sure if this is an onion or how you know so much about onions but I really appreciate your help and maintaining a working example.
Hopefully my ājob succeededā wasnāt a fluke and I dont have to come back here with more head scratching ;o)
1 Like
@mariusrugan - one more question - not 100% on terminology - but is there a way with gitea act runners using this DinD configuration to use an in-cluster mirror. (k3s for example calls this an embedded mirror).
Iām also interested in a docker layer cache that is shared amongst the runners and not having to rebuild everything from scratch. Iāve noticed, in cluster, with this new shiny DinD setup that every workflow job rebuilds a Dockerfile from scratch as there is no existing layers from previous builds.
Any thoughts on how to configure this or if its even possible to get act runner to use a registry mirror or build cache?