ci.yml

name: CI

on:
  push:
    branches: [main, develop, 'feature/**', 'bugfix/**']
  pull_request:
    branches: [main, develop]
  workflow_dispatch:

permissions:
  contents: read
  security-events: write

jobs:
  validate:
    name: Validate Configuration
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Validate Docker Compose files
        run: |
          echo "Validating: docker/docker-compose.standalone.yml"
          docker compose -f docker/docker-compose.standalone.yml config -q
          echo "OK: docker/docker-compose.standalone.yml"
          echo "Validating: docker/docker-compose.lan.yml"
          docker compose -f docker/docker-compose.lan.yml config -q
          echo "OK: docker/docker-compose.lan.yml"
          for f in docker/docker-compose.advanced.yml docker/docker-compose.sso.yml; do
            echo "Checking scaffold: $f"
            docker compose -f "$f" config --no-interpolate -q 2>&1 && echo "OK: $f" \
              || echo "WARN: $f has placeholder variables (scaffold — not yet built out)"
          done
          echo "Validating: docker/docker-compose.integration.yml"
          docker compose -f docker/docker-compose.integration.yml config -q
          echo "OK: docker/docker-compose.integration.yml"
          echo "Validating: docker/docker-compose.production.yml"
          docker compose -f docker/docker-compose.production.yml config -q
          echo "OK: docker/docker-compose.production.yml"

      - name: ShellCheck — lab test scripts
        run: |
          sudo apt-get install -y shellcheck -qq
          shellcheck --severity=error tests/labs/*.sh

      - name: Validate module manifest
        run: |
          python3 -c "
          import sys, re
          with open('it-stack-traefik.yml') as f:
              content = f.read()
          required = ['module:', 'version:', 'phase:', 'category:', 'ports:']
          missing = [k for k in required if k not in content]
          if missing:
              print('Missing fields:', missing); sys.exit(1)
          print('Manifest valid')
          "

  security-scan:
    name: Security Scan
    runs-on: ubuntu-latest
    needs: validate
    steps:
      - uses: actions/checkout@v4

      - name: Trivy — scan Dockerfile
        uses: aquasecurity/trivy-action@0.28.0
        with:
          scan-type: config
          scan-ref: .
          exit-code: '0'
          severity: CRITICAL,HIGH

      - name: Trivy — SARIF output
        uses: aquasecurity/trivy-action@0.28.0
        with:
          scan-type: config
          scan-ref: .
          format: sarif
          output: trivy-results.sarif

      - name: Upload SARIF to GitHub Security
        uses: github/codeql-action/upload-sarif@v3
        if: always()
        with:
          sarif_file: trivy-results.sarif

  lab-01-smoke:
    name: Lab 01 — Smoke Test
    runs-on: ubuntu-latest
    needs: validate
    continue-on-error: true   # scaffold stubs; full lab runs on real VMs
    steps:
      - uses: actions/checkout@v4

      - name: Install tools
        run: sudo apt-get install -y netcat-openbsd

      - name: Start standalone stack
        run: docker compose -f docker/docker-compose.standalone.yml up -d

      - name: Wait for Traefik to be ready
        run: |
          echo "Waiting for Traefik ping endpoint..."
          timeout 60 bash -c 'until curl -sf http://localhost:80/ping > /dev/null 2>&1; do sleep 2; done'
          echo "Traefik is ready"
          docker compose -f docker/docker-compose.standalone.yml ps

      - name: Run Lab 18-01 test script
        run: bash tests/labs/test-lab-18-01.sh

      - name: Collect logs on failure
        if: failure()
        run: docker compose -f docker/docker-compose.standalone.yml logs

      - name: Cleanup
        if: always()
        run: docker compose -f docker/docker-compose.standalone.yml down -v

  lab-02-smoke:
    name: Lab 02 — TLS Routing & Load Balancing
    runs-on: ubuntu-latest
    needs: validate
    continue-on-error: true
    steps:
      - uses: actions/checkout@v4

      - name: Install tools
        run: sudo apt-get install -y netcat-openbsd curl

      - name: Start LAN stack (traefik + backends)
        run: docker compose -f docker/docker-compose.lan.yml up -d

      - name: Wait for Traefik ping endpoint
        run: |
          timeout 90 bash -c '
            until curl -sf http://localhost:80/ping; do
              sleep 3
            done'
          echo "Traefik is ready"

      - name: Run Lab 18-02 test script
        run: bash tests/labs/test-lab-18-02.sh

      - name: Collect logs on failure
        if: failure()
        run: docker compose -f docker/docker-compose.lan.yml logs

      - name: Cleanup
        if: always()
        run: docker compose -f docker/docker-compose.lan.yml down -v

  lab-03-smoke:
    name: Lab 03 — Prometheus Metrics + Middleware + TCP
    runs-on: ubuntu-latest
    needs: validate
    continue-on-error: true
    steps:
      - uses: actions/checkout@v4

      - name: Install tools
        run: sudo apt-get install -y netcat-openbsd curl

      - name: Start advanced stack (traefik + backends)
        run: docker compose -f docker/docker-compose.advanced.yml up -d

      - name: Wait for Traefik ping
        run: timeout 90 bash -c 'until curl -sf http://localhost:80/ping; do sleep 3; done'

      - name: Wait for metrics endpoint
        run: timeout 30 bash -c 'until curl -sf http://localhost:8082/metrics | grep -q traefik_; do sleep 3; done'

      - name: Run Lab 18-03 test script
        run: bash tests/labs/test-lab-18-03.sh

      - name: Collect logs on failure
        if: failure()
        run: docker compose -f docker/docker-compose.advanced.yml logs

      - name: Cleanup
        if: always()
        run: docker compose -f docker/docker-compose.advanced.yml down -v

  lab-04-smoke:
    name: Lab 04 — Traefik ForwardAuth SSO via Keycloak
    runs-on: ubuntu-latest
    needs: validate
    continue-on-error: true
    steps:
      - uses: actions/checkout@v4

      - name: Install tools
        run: sudo apt-get install -y netcat-openbsd curl

      - name: Start SSO stack (KC + Traefik + oauth2-proxy + whoami services)
        run: docker compose -f docker/docker-compose.sso.yml up -d

      - name: Wait for Keycloak
        run: timeout 200 bash -c 'until curl -sf http://localhost:8080/health/ready | grep -q UP; do sleep 5; done'

      - name: Wait for Traefik
        run: timeout 60 bash -c 'until curl -sf http://localhost:8080/api/version | grep -q Version; do sleep 3; done'

      - name: Run Lab 18-04 test script
        env:
          KC_PASS: "Lab04Password!"
        run: bash tests/labs/test-lab-18-04.sh

      - name: Collect logs on failure
        if: failure()
        run: docker compose -f docker/docker-compose.sso.yml logs

      - name: Cleanup
        if: always()
        run: docker compose -f docker/docker-compose.sso.yml down -v

  lab-05-smoke:
    name: Lab 05 -- Traefik + Keycloak + ForwardAuth + Prometheus
    runs-on: ubuntu-latest
    needs: validate
    continue-on-error: true
    steps:
      - uses: actions/checkout@v4

      - name: Install tools
        run: sudo apt-get install -y curl netcat-openbsd

      - name: Start integration stack
        run: docker compose -f docker/docker-compose.integration.yml up -d

      - name: Wait for Keycloak
        run: timeout 200 bash -c 'until curl -sf http://localhost:8080/health/ready | grep -q UP; do sleep 5; done'

      - name: Wait for Traefik
        run: timeout 60 bash -c 'until curl -sf http://localhost:8080/api/version | grep -q Version; do sleep 3; done'

      - name: Run Lab 18-05 test script
        env:
          KC_PASS: "Lab05Password!"
        run: bash tests/labs/test-lab-18-05.sh

      - name: Collect logs on failure
        if: failure()
        run: docker compose -f docker/docker-compose.integration.yml logs

      - name: Cleanup
        if: always()
        run: docker compose -f docker/docker-compose.integration.yml down -v

  lab-06-smoke:
    name: Lab 06 — Traefik Production (TLS, rate-limit, access logs, Prometheus)
    runs-on: ubuntu-latest
    needs: validate
    continue-on-error: true
    steps:
      - uses: actions/checkout@v4

      - name: Install tools
        run: sudo apt-get install -y curl

      - name: Start production stack
        run: docker compose -f docker/docker-compose.production.yml up -d

      - name: Wait for Traefik API
        run: timeout 60 bash -c 'until curl -sf http://localhost:8080/api/version | grep -q Version; do sleep 3; done'

      - name: Wait for backends
        run: timeout 60 bash -c 'until curl -sf http://localhost/ | grep -q Hostname; do sleep 3; done'

      - name: Run Lab 18-06 test script
        run: bash tests/labs/test-lab-18-06.sh

      - name: Collect logs on failure
        if: failure()
        run: docker compose -f docker/docker-compose.production.yml logs

      - name: Cleanup
        if: always()
        run: docker compose -f docker/docker-compose.production.yml down -v