Plugin implementing Gitlab Fleeting for use with Kubevirt VM's.

1. Get Kubernetes cluster with working Kubevirt
2. Create namespace

   kubectl create namespace gitlab-kubevirt-test
   

3. Generate SSH key pair for automated instance access (e. g. using ssh-keygen) and put the result into a secret:

   $ ssh-keygen -f ssh_id_fleeting -N ''
   ...
   $ kubectl create secret generic -n gitlab-kubevirt-test ssh-key --from-file=ssh-privatekey=ssh_id_fleeting
   

4. Add values.yaml for Gitlab runner helm chart (See docs). Reference values.yaml for full docker-autoscaler example (simpler "instance" executor should work as well), note comments; refer to helm chart docs for use of general parameters:

   gitlabUrl: https://my-gitlab.domain.invalid/ # Replace
   runnerToken: "...token..." # Replace
   certsSecretName: "xyz-certs" # Replace, if any custom certs
   concurrent: 3
   preEntrypointScript: |
     gitlab-runner fleeting install
   serviceAccount:
     create: true
   rbac:
     create: true
     rules:
       - apiGroups: ["kubevirt.io"]
         resources: ["virtualmachines", "virtualmachineinstances"]
         verbs: ["get", "list", "create", "delete"]
   
   runners:
     executor: "docker-autoscaler"
     config: |
       [[runners]]
         request_concurrency = 2
   
         [runners.docker]
           image = "alpine:latest"
   
         [runners.autoscaler]
           plugin = "ghcr.io/gonicus/gitlab-fleeting-plugin-kubevirt:0.0.13" # Or build your own
           capacity_per_instance = 1
           max_use_count = 1
           max_instances = 9
           instance_ready_command = "cloud-init status --wait"
           update_interval_when_expecting = "30s"
   
         [runners.autoscaler.plugin_config]
           useInClusterConfig = true
           vmLabelKey = "type"
           vmLabelValue = "gitlabfleet"
           vmNamespace = "gitlab-kubevirt-test" # Be sure to set to the same namespace as the helm release
           vmNamePrefix = "gitlab-fleet-vm"
           vmReadinessProbeScript = "cat /tmp/healthy.txt"
           vmRAM = "4Gi"
           vmCPUCores = "4"
           vmRunnerImage = "quay.io/containerdisks/debian:13" # Or - preferably - build your own with all required tooling baked in
           vmCloudInitUserData = '''
       #cloud-config
       ssh_pwauth: False
       ssh_authorized_keys:
         - 'ssh-ed25519 ...' # Put in content of ssh_id_fleeting.pub
       runcmd:
         - apt update && apt -y install curl git qemu-guest-agent ca-certificates docker.io
         - |
             # Add custom CA certificate, if any
             cat > /usr/local/share/ca-certificates/CA.crt <<EOF
             -----BEGIN CERTIFICATE-----
             ...
             -----END CERTIFICATE-----
             EOF
         - update-ca-certificates
         - usermod -aG docker debian
         - systemctl enable --now docker.service
         - touch /tmp/healthy.txt
         - systemctl enable --now qemu-guest-agent'''
   
         [runners.autoscaler.connector_config]
           username               = "debian"
           key_path               = "/secrets/ssh-privatekey"
           use_static_credentials = true
           timeout                = "5m0s"
           use_external_addr      = false
   
         [[runners.autoscaler.policy]]
           idle_count = 3
           idle_time  = "20m0s"
           preemptive_mode = true
   secrets:
     - name: "ssh-key"
       items:
         - key: "ssh-privatekey"
           path: "ssh-privatekey"
   

5. Install using helm

   helm repo add gitlab https://charts.gitlab.io
   helm install --namespace gitlab-kubevirt-test gitlab-runner -f values.yaml gitlab/gitlab-runner