runner-heartbeat.yml

name: Runner Heartbeat

on:
  schedule:
    - cron: '0 0 */7 * *'  # Every 7 days at midnight UTC
  workflow_dispatch:

concurrency:
  group: runner-heartbeat
  cancel-in-progress: true

permissions:
  contents: read

env:
  GCP_PROJECT_ID: ${{ secrets.GCP_PROJECT_ID }}

jobs:
  start-builder:
    name: Start areal-docker-builder instance
    runs-on: ubuntu-latest
    env:
      INSTANCE_NAME: areal-docker-builder
      INSTANCE_ZONE: us-central1-f
    steps:
      - name: Authenticate to Google Cloud
        uses: google-github-actions/auth@v3
        with:
          credentials_json: ${{ secrets.GCP_SERVICE_ACCOUNT_KEY }}

      - name: Set up Google Cloud SDK
        uses: google-github-actions/setup-gcloud@v3

      - name: Start builder instance if stopped
        run: |
          set -euo pipefail

          status=$(gcloud compute instances describe "$INSTANCE_NAME" \
            --project "$GCP_PROJECT_ID" \
            --zone "$INSTANCE_ZONE" \
            --format="get(status)" || echo "NOT_FOUND")

          if [ "$status" = "NOT_FOUND" ]; then
            echo "Error: Instance $INSTANCE_NAME not found in zone $INSTANCE_ZONE" >&2
            exit 1
          fi

          if [ "$status" = "RUNNING" ]; then
            echo "Instance $INSTANCE_NAME is already running."
          elif [ "$status" = "TERMINATED" ] || [ "$status" = "SUSPENDED" ]; then
            echo "Instance $INSTANCE_NAME is $status. Starting it..."
            gcloud compute instances start "$INSTANCE_NAME" \
              --project "$GCP_PROJECT_ID" \
              --zone "$INSTANCE_ZONE"
            echo "Instance started successfully."
          else
            echo "Instance $INSTANCE_NAME has unexpected status: $status" >&2
            exit 1
          fi

      - name: Wait for builder runner to be online
        uses: actions/github-script@v8
        env:
          INSTANCE_NAME: areal-docker-builder
          GH_PAT: ${{ secrets.GH_PAT }}
        with:
          github-token: ${{ secrets.GH_PAT }}
          script: |
            const instanceName = process.env.INSTANCE_NAME;
            const maxAttempts = 120;
            const delayMs = 10000;

            const pat = process.env.GH_PAT;
            if (!pat) {
              core.setFailed('GH_PAT secret is not configured.');
              return;
            }

            const wait = (ms) => new Promise((resolve) => setTimeout(resolve, ms));

            for (let attempt = 1; attempt <= maxAttempts; attempt++) {
              const response = await github.rest.actions.listSelfHostedRunnersForRepo({
                owner: context.repo.owner,
                repo: context.repo.repo,
                per_page: 100,
                request: {
                  headers: {
                    authorization: `token ${pat}`,
                  },
                },
              });

              const found = response.data.runners.find((runner) => runner.name === instanceName);
              if (found && found.status === 'online') {
                core.info(`Builder runner ${instanceName} is online.`);
                return;
              }

              core.info(`Builder runner ${instanceName} not ready yet (attempt ${attempt}/${maxAttempts}).`);
              await wait(delayMs);
            }

            throw new Error(`Timed out waiting for builder runner ${instanceName} to come online.`);

  heartbeat:
    name: Runner heartbeat
    needs: start-builder
    runs-on: [self-hosted, areal-docker-builder]
    steps:
      - name: Heartbeat
        run: |
          echo "Runner heartbeat at $(date -u)"
          echo "Runner: $(hostname)"
          echo "Uptime: $(uptime)"

  stop-builder:
    name: Stop areal-docker-builder instance
    needs:
      - heartbeat
      - start-builder
    if: always()
    runs-on: ubuntu-latest
    env:
      INSTANCE_NAME: areal-docker-builder
      INSTANCE_ZONE: us-central1-f
    steps:
      - name: Authenticate to Google Cloud
        uses: google-github-actions/auth@v3
        with:
          credentials_json: ${{ secrets.GCP_SERVICE_ACCOUNT_KEY }}

      - name: Set up Google Cloud SDK
        uses: google-github-actions/setup-gcloud@v3

      - name: Stop builder instance
        run: |
          status=$(gcloud compute instances describe "$INSTANCE_NAME" \
            --project "$GCP_PROJECT_ID" \
            --zone "$INSTANCE_ZONE" \
            --format="get(status)" 2>/dev/null || echo "NOT_FOUND")

          if [ "$status" = "NOT_FOUND" ]; then
            echo "Warning: Instance $INSTANCE_NAME not found in zone $INSTANCE_ZONE"
            exit 0
          fi

          if [ "$status" = "RUNNING" ]; then
            echo "Stopping instance $INSTANCE_NAME..."
            gcloud compute instances stop "$INSTANCE_NAME" \
              --project "$GCP_PROJECT_ID" \
              --zone "$INSTANCE_ZONE"
            echo "Instance stopped successfully."
          else
            echo "Instance $INSTANCE_NAME is already in status: $status"
          fi