home-ops/.github/workflows/template-e2e-cluster.yaml

---
# yaml-language-server: $schema=https://json.schemastore.org/github-workflow.json
name: "E2E Cluster"

on:
  workflow_dispatch:
  pull_request:
    branches: ["main"]
  schedule:
    - cron: "30 5 * * *"

concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

permissions: {}

jobs:
  bootstrap:
    if: ${{ github.repository == 'onedr0p/cluster-template' }}
    name: bootstrap (qemu)
    runs-on: ubuntu-latest
    permissions:
      contents: read
    services:
      git:
        image: ghcr.io/charmbracelet/soft-serve:v0.12.2
        env:
          SOFT_SERVE_GIT_ENABLED: "false"
          SOFT_SERVE_LFS_ENABLED: "false"
          SOFT_SERVE_SSH_LISTEN_ADDR: "127.0.0.1:23231"
          SOFT_SERVE_STATS_ENABLED: "false"
        ports:
          - 8418:23232
        entrypoint: /bin/sh
        command: >-
          -c "set -eu; ssh-keygen -q -t ed25519 -N '' -f /tmp/admin;
          export SOFT_SERVE_INITIAL_ADMIN_KEYS=$(cat /tmp/admin.pub);
          /usr/local/bin/soft serve & pid=$!;
          until ssh -q -i /tmp/admin -o IdentitiesOnly=yes -o StrictHostKeyChecking=no -p 23231 localhost settings anon-access read-write; do sleep 1; done;
          ssh -q -i /tmp/admin -o IdentitiesOnly=yes -o StrictHostKeyChecking=no -p 23231 localhost repo create repo;
          wait $pid"
        options: >-
          --health-cmd "git ls-remote http://localhost:23232/repo.git"
          --health-interval 2s
          --health-timeout 2s
          --health-retries 30
    steps:
      - name: Checkout
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false

      - name: Install QEMU
        run: |
          sudo apt-get update
          sudo apt-get install --yes --no-install-recommends gettext-base qemu-system-x86 qemu-utils ovmf

      - name: Setup mise
        uses: jdx/mise-action@c2a87611a18de5b3828c5652fe268e992400cb5c # v4.3.0
        env:
          GITHUB_TOKEN: "${{ secrets.GITHUB_TOKEN }}"

      - name: Boot maintenance-mode nodes
        id: cluster
        uses: home-operations/talosctl-cluster-action@fb6a31bf5de43218acc80d2e23958a16eee7380c # v0.2.2
        with:
          config: ./.github/template-tests/e2e/talos-cluster.yaml
          cache: true

      - name: Export cluster environment
        env:
          CONTROLPLANE_IPS: "${{ steps.cluster.outputs.controlplane-ips }}"
          GATEWAY: "${{ steps.cluster.outputs.gateway }}"
          WORKER_IPS: "${{ steps.cluster.outputs.worker-ips }}"
        run: |
          echo "E2E_CONTROLPLANE_IPS=$CONTROLPLANE_IPS" >> "$GITHUB_ENV"
          echo "E2E_WORKER_IPS=$WORKER_IPS" >> "$GITHUB_ENV"
          echo "E2E_GATEWAY=$GATEWAY" >> "$GITHUB_ENV"
          echo "E2E_CIDR=10.9.0.0/24" >> "$GITHUB_ENV"
          echo "E2E_STATE=$RUNNER_TEMP/template-e2e" >> "$GITHUB_ENV"

      - name: Prepare cluster
        run: bash ./.github/template-tests/e2e/cluster.sh prepare

      - name: Build healthy cluster foundation
        run: bash ./.github/template-tests/e2e/cluster.sh foundation

      - name: Test Flux SOPS
        id: flux-sops
        run: bash ./.github/template-tests/e2e/cluster.sh flux-sops
        background: true

      - name: Test networking
        id: networking
        run: bash ./.github/template-tests/e2e/cluster.sh networking
        background: true

      - name: Wait for E2E tests
        wait: [flux-sops, networking]

      - name: Summarize cluster
        run: bash ./.github/template-tests/e2e/cluster.sh summary