test-environment-validation.yaml

name: Test Environment Setup and Validation

on:
  push:
    branches:
      - main
  workflow_dispatch:
    inputs:
      test_suite:
        description: 'Test suite to run (e.g., resourceindexpolicy, or empty for all)'
        required: false
        default: ''
        type: string
  pull_request: {}

env:
  # Enable experimental remote taskfiles feature
  TASK_X_REMOTE_TASKFILES: 1
  # Test infrastructure configuration
  TEST_INFRA_CLUSTER_NAME: test-infra
  SEARCH_IMAGE_NAME: ghcr.io/datum-cloud/search
  SEARCH_IMAGE_TAG: dev

jobs:
  test-environment-validation:
    runs-on: ubuntu-latest
    timeout-minutes: 60

    steps:
      - name: Checkout code
        uses: actions/checkout@v6

      - name: Set up Go
        uses: actions/setup-go@v6
        with:
          go-version-file: 'go.mod'
          cache: true

      - name: Install Task CLI
        run: |
          sh -c "$(curl --location https://taskfile.dev/install.sh)" -- -d -b /usr/local/bin

      - name: Verify Task installation
        run: |
          task --version
          echo "Available tasks:"
          task --list

      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v4
        with:
          buildkitd-config-inline: |
            [worker.oci]
              max-parallelism = 4

      - name: Install kubectl
        uses: azure/setup-kubectl@v4
        with:
          version: 'v1.30.0'

      - name: Install KinD
        uses: helm/kind-action@v1
        with:
          install_only: true
          version: v0.24.0

      - name: Verify prerequisites
        run: |
          echo "=== Checking prerequisites ==="
          docker version
          kubectl version --client
          kind version
          echo "Go version: $(go version)"

      - name: Set up test environment
        run: |
          echo "=== Setting up test environment ==="
          task ci:setup

      - name: Verify Search components
        run: |
          echo "=== Verifying Search components ==="

          # Verify all Search components are running
          echo "Checking Search control plane components:"
          task test-infra:kubectl -- get pods -n search-system

          # Wait for controller manager to be ready
          echo "⏳ Waiting for controller manager to be ready..."
          task test-infra:kubectl -- wait --for=condition=Ready pod -l app.kubernetes.io/name=search-controller-manager -n search-system --timeout=1000s

          # Wait for API server to be ready
          echo "⏳ Waiting for API server to be ready..."
          task test-infra:kubectl -- wait --for=condition=Ready pod -l app.kubernetes.io/name=search-apiserver -n search-system --timeout=1000s

          # Verify Aggregated API Availability (CA Injection)
          echo "⏳ Verifying Aggregated API Availability..."
          for i in {1..30}; do
            CA_LEN=$(task test-infra:kubectl -- get apiservice v1alpha1.search.miloapis.com -o jsonpath='{len(.spec.caBundle)}' 2>/dev/null || echo "0")
            if [ "$CA_LEN" -gt "0" ]; then
              echo "✅ CA Bundle injected into APIService."
              break
            fi
            echo "Waiting for CA injection into APIService..."
            sleep 2
          done
          
          # Verify Discovery works
          echo "Verifying API Discovery..."
          task test-infra:kubectl -- get resourceindexpolicies
          
          echo "✓ Search components verification complete"

      - name: Run end-to-end tests
        run: |
          echo "=== Running end-to-end tests ==="

          # Determine which tests to run based on input
          if [ -n "${{ github.event.inputs.test_suite }}" ]; then
            echo "Running specified test suite: ${{ github.event.inputs.test_suite }}"
            task test:end-to-end -- ${{ github.event.inputs.test_suite }}
          else
            echo "Running all end-to-end tests..."
            task test:end-to-end
          fi

      - name: Collect debug information on failure
        if: failure()
        run: |
          echo "=== Collecting debug information ==="

          # Cluster status
          echo "=== Infrastructure Cluster Status ==="
          task test-infra:kubectl -- get pods -A || true
          task test-infra:kubectl -- get nodes -o wide || true

          # Search control plane status
          echo "=== Search Control Plane Status ==="
          task test-infra:kubectl -- describe pods -n search-system || true

          # Search API server logs
          echo "=== Search API Server Logs ==="
          task test-infra:kubectl -- logs -n search-system -l app.kubernetes.io/name=search-apiserver --tail=100 || true

          # Controller manager logs
          echo "=== Controller Manager Logs ==="
          task test-infra:kubectl -- logs -n search-system -l app.kubernetes.io/name=search-controller-manager --tail=100 || true

          # Docker container status
          echo "=== Docker Containers ==="
          docker ps -a || true

          # KinD cluster info
          echo "=== KinD cluster info ==="
          kind get clusters || true
          kind export logs /tmp/kind-logs --name $TEST_INFRA_CLUSTER_NAME || true

      - name: Upload debug artifacts
        if: failure()
        uses: actions/upload-artifact@v7
        with:
          name: debug-logs
          path: |
            /tmp/kind-logs/
          if-no-files-found: ignore

      - name: Cleanup test infrastructure
        if: always()
        run: |
          echo "=== Cleaning up test infrastructure ==="

          # Clean up test infrastructure cluster
          task test-infra:cluster-down || true

          # Verify cleanup
          echo "Remaining KinD clusters:"
          kind get clusters || true

          echo "Remaining Docker containers:"
          docker ps -a --filter "name=test-infra" || true