GitHub Actions Runner Controller (ARC) - Interview Architecture

Interview-Ready Architecture Diagrams for AMD Technical Interview

Diagram 1: Simple High-Level Architecture

Talk Track for Interview:

“GitHub ARC runs self-hosted runners on Kubernetes. When a workflow is triggered, GitHub sends the job to the ARC Controller via a GitHub App. The Controller communicates with the Runner Scale Set, which creates ephemeral runner pods to execute jobs. After completion, the pods are destroyed-ensuring clean, secure environments for every build.”

Diagram 2: Scaling Behavior

Talk Track for Interview:

“ARC implements scale-to-zero architecture. At rest, only the controller runs-no runner pods consuming resources. When jobs arrive, it scales instantly from 0 to N runners. After jobs complete, runners terminate automatically. This saves costs compared to static Jenkins agents that run 24/7.”

Diagram 3: Multi-Environment Setup

Talk Track for Interview:

“For AMD’s scale, I’d recommend multiple ARC deployments across environments. Each cluster runs its own controller with labeled runner sets. Workflows select runners using runs-on: [staging, test] labels. This isolates production from development and allows per-environment security policies and resource quotas.”

Diagram 4: ARC vs Jenkins Comparison

Talk Track for Interview:

“Moving from Jenkins to ARC solves AMD’s key pain points:

• Cost: Jenkins agents run 24/7. ARC scales to zero = 70% infrastructure savings.
• Security: Shared Jenkins agents retain state. ARC pods are ephemeral = fresh environment every job.
• Maintenance: Jenkins requires plugin updates, agent patching. ARC auto-updates via Helm.
• Speed: Jenkins takes minutes to provision agents. ARC pods start in 10-20 seconds.”

Deployment Commands (For Interview Demo)

# Install ARC Controller via Helm
helm install arc-controller \
  oci://ghcr.io/actions/actions-runner-controller-charts/gha-runner-scale-set-controller \
  --namespace arc-systems \
  --create-namespace
 
# Deploy Runner Scale Set
helm install arc-runners \
  oci://ghcr.io/actions/actions-runner-controller-charts/gha-runner-scale-set \
  --namespace arc-runners \
  --set githubConfigUrl="https://github.com/AMD" \
  --set githubConfigSecret=github-app-secret \
  --set minRunners=0 \
  --set maxRunners=100

Talk Track:

“ARC deploys via Helm in under 5 minutes. I’d use Terraform to manage this across AMD’s EKS clusters, with environment-specific values for dev/staging/prod. The GitHub App secret authenticates the controller-no PATs needed.”

Key Interview Stats to Memorize

Sample Workflow Using ARC Runners

name: Build on ARC
on: push
 
jobs:
  build:
    runs-on: [self-hosted, amd-production]  # Uses ARC runners
    steps:
      - uses: actions/checkout@v4
      - name: Build
        run: npm run build
      - name: Test
        run: npm test

Talk Track:

“Workflows use runs-on: [self-hosted, amd-production] to target ARC runners. The controller sees the job, spins up a pod with the amd-production label, executes the workflow, and terminates the pod. Developers don’t change their workflow code-just the runs-on label.”

Interview Opening Line (Use This!)

“I recommend deploying GitHub Actions Runner Controller on AMD’s existing Kubernetes infrastructure. It reduces runner costs by 70% through scale-to-zero, improves security with ephemeral pods, and integrates seamlessly with GitHub Actions Importer to migrate 500 pipelines in under 6 months instead of 2 years.”