GitHub Actions - Self-hosted Linux Runners

These folders contain the required resources to deploy the GitHub Actions self-hosted runners for the workflows running in Ubuntu OS.

• /docker

  • Dockerfile and entrypoint: Resources needed to create a new self-hosted-runner image.
  • docker-compose.yml: In case you would like to test and run the self-hosted runner locally.

• /kubernetes

  • Kubernetes files to create the resources needed to deploy the self-hosted runners.

Docker

How to build a new image and push it to the Artifact Registry?

• Create the image locally

docker build -t $LOCAL_IMAGE_NAME:TAG .

• Tag the local image with the GCP repository name

docker tag $LOCAL_IMAGE_NAME LOCATION-docker.pkg.dev/PROJECT-ID/REPOSITORY/IMAGE:latest

• Make sure you are authenticated to the gcloud repository

gcloud auth configure-docker us-central1-docker.pkg.dev

• Push the tagged image to the repository

docker push LOCATION-docker.pkg.dev/PROJECT-ID/REPOSITORY/IMAGE:latest

• Important: Please add the commit hash as a tag when a new image is uploaded.

How to run a self-hosted runner locally?

• Create a .var.env file from the example.var.env file and replace the corresponding values:

  • GOOGLE_APPLICATION_CREDENTIALS: Path where the json file will be stored (Same path as the one given in the docker-compose.yml file)
  • CLOUD_FUNCTION_NAME: Function where the self-hosted runner will get the RUNNER_TOKEN
  • ORG_RUNNER_GROUP: Name of the GitHub Actions Runner Group. You can find it here: https://github.com/organizations/ORG_NAME/settings/actions/runner-groups
  • ORG_NAME: Name of the Organization
  • GCP_REGION: GCP Region where your Cloud Function is deployed
  • GCP_PROJECT_ID: GCP Project ID where your Cloud Function is deployed

• Run the container

docker compose up

• You will be able to see the self-hosted runner in the Settings of your GitHub repository

Kubernetes

For the current implementation we are using Google Kubernetes Engine (GKE) as a container orchestration platform.

• Configure your Kubernetes local context

gcloud container clusters get-credentials $GCP_CLUSTER_NAME --zone $GCP_REGION --project $GCP_PROJECT-ID

• Create the desired k8s namespace

kubectl create namespace $NAMESPACE

• Create the GKE secret from a json file

kubectl create secret generic $k8s_SECRET_NAME --from-file=key.json=$LOCAL_PATH --namespace $NAMESPACE

• Update the github-actions-secrets.yml file with its corresponding values encrypted in base64

echo -n "$VARIABLE_VALUE" | base64

• Replace in github-actions-deployment.yml file the $IMAGE_URL variable with the corresponding image URL: GCP_LOCATION-docker.pkg.dev/GCP_PROJECT_ID/REPOSITORY_NAME/IMAGE_NAME

• In case you would like to create the deployment from scratch, run the run-k8s-deployment.sh script to execute the Kubernetes deployment in the GKE cluster.

  • Important: Make sure you have the GKE context selected in your local machine: kubectl config current-context

./run-k8s-deployment.sh $NAMESPACE

• Otherwise, apply only the changes. Important: Make sure you have the GKE context selected in your local machine: kubectl config current-context

kubectl apply -f github-actions-$FILE_NAME.yml --namespace $NAMESPACE

• In case you would like to delete all the Kubernetes resources, run the delete-k8s-deployment.sh script with its corresponding namespace value.

./delete-k8s-deployment.sh $NAMESPACE