ci.yml

name: CI

on:
  push:
    branches:
      - main
      - release-*
  workflow_dispatch: {}
  pull_request:
    branches:
      - main
      - release-*
    paths-ignore: [docs/**, "**.md", "**.mdx", "**.png", "**.jpg"]

env:
  GO_VERSION: '1.25.10'
  CERT_MANAGER_VERSION: 'v1.16.2'

jobs:
  detect-noop:
    runs-on: ubuntu-latest
    outputs:
      noop: ${{ steps.noop.outputs.should_skip }}
    steps:
      - name: Detect No-op Changes
        id: noop
        uses: fkirc/skip-duplicate-actions@f75f66ce1886f00957d99748a42c724f4330bdcf # v5.3.1
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}
          do_not_skip: '["workflow_dispatch", "schedule", "push"]'
          concurrent_skipping: false

  unit-and-integration-tests:
    runs-on: ubuntu-latest
    needs: detect-noop
    if: needs.detect-noop.outputs.noop != 'true'
    steps:
      - name: Set up Go
        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
        with:
          go-version: ${{ env.GO_VERSION }}

      - name: Check out code into the Go module directory
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
              
      - name: Set up Ginkgo CLI
        run: |
          go install github.com/onsi/ginkgo/v2/ginkgo@v2.19.1
      
      - name: Prepare necessary environment variables
        run: |
          echo "CGO_ENABLED=1" >> $GITHUB_ENV
          KUBEBUILDER_ASSETS=$(make --silent kubebuilder-assets-path)
          echo "KUBEBUILDER_ASSETS="$KUBEBUILDER_ASSETS"" >> $GITHUB_ENV

      # Certain tests that require special setup (e.g., those that should be run with Ginkgo CLI only) will
      # be skipped in this step.
      #
      # Note that the skipping only applies to the CI environment.
      - name: Run unit and integration tests with default setup & generate coverage
        run: |
          make test
        env:
          KUBEFLEET_CI_TEST_RUNNER_NAME: 'default'
      
      # The work applier integration tests use in-memory Kubernetes environment setup; due to resource constraints
      # and the way the tests are organized, running the suite with as many parallel Ginkgo processes as possible (i.e.,
      # the number of all CPU cores) might not lead to the optimal outcome.
      #
      # Note (chenyu1): switch to test matrices if we need to test with more configuration combos in the future.
      - name: Run work applier unit and integration tests with Ginkgo CLI & generate coverage
        run: |
          ginkgo -v -p --procs=4 --race --cover -coverprofile=work-applier-it-coverage.out ./pkg/controllers/workapplier/ 
          KUBEFLEET_CI_WORK_APPLIER_RUN_WITH_PRIORITY_QUEUE=true ginkgo -v -p --procs=4 --race --cover -coverprofile=work-applier-it-no-pri-q-coverage.out ./pkg/controllers/workapplier/
        env:
          KUBEFLEET_CI_TEST_RUNNER_NAME: 'ginkgo'
      
      - name: Upload Codecov report
        uses: codecov/codecov-action@57e3a136b779b570ffcdbf80b3bdc90e7fab3de2 # v6.0.0
        with:
          ## Repository upload token - get it from codecov.io. Required only for private repositories
          token: ${{ secrets.CODECOV_TOKEN }}
          # The codecov action will auto-search all coverage files by default. All uploaded coverage will be
          # merged automatically.

  e2e-tests:
    strategy:
      fail-fast: false
      matrix:
        customized-settings: [default, resourceplacement, joinleave, custom]
        include:
          - customized-settings: default
            # to shorten the test duration, set the resource snapshot creation interval to 0
            resource-snapshot-creation-minimum-interval: 0m
            resource-changes-collection-duration: 0m
          - customized-settings: resourceplacement
            # to shorten the test duration, set the resource snapshot creation interval to 0
            resource-snapshot-creation-minimum-interval: 0m
            resource-changes-collection-duration: 0m
          - customized-settings: joinleave
            # to shorten the test duration, set the resource snapshot creation interval to 0
            resource-snapshot-creation-minimum-interval: 0m
            resource-changes-collection-duration: 0m
          - customized-settings: custom
            resource-snapshot-creation-minimum-interval: 30s
            resource-changes-collection-duration: 15s
    runs-on: ubuntu-latest
    needs: [
      detect-noop,
    ]
    if: needs.detect-noop.outputs.noop != 'true'
    steps:
      - name: Set up Go
        uses: actions/setup-go@4a3601121dd01d1626a1e23e37211e3254c1c06c # v6
        with:
          go-version: ${{ env.GO_VERSION }}

      - name: Check out code into the Go module directory
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
      
      - name: Move Docker data directory to /mnt
        # The default storage device on GitHub-hosted runners is running low during e2e tests.
        # Moving Docker data directory to /mnt which has more space and is backed by a larger
        # storage device.
        # Upstream kubefleet repo has a much bigger default storage device and no secondary
        # storage device to use.
        run: |
          echo "=== Moving Docker to /mnt for more disk space ==="
          df -h
          sudo systemctl stop docker
          echo '{"data-root": "/mnt/docker"}' | sudo tee /etc/docker/daemon.json
          sudo mkdir -p /mnt/docker
          if [ -d "/var/lib/docker" ]; then
            sudo mv /var/lib/docker/* /mnt/docker/ || true
          fi
          sudo systemctl start docker
          echo "=== Docker moved to /mnt, verifying ==="
          docker info | grep "Docker Root Dir" || true

      - name: Install Ginkgo CLI
        run: |
          go install github.com/onsi/ginkgo/v2/ginkgo@v2.23.4
      
      - name: Install Kind
        # Before updating the kind version to use, verify that the current kind image
        # is still supported by the version.
        run: |
          go install sigs.k8s.io/kind@v0.30.0

      - name: Run e2e tests
        run: |
          if [ "${{ matrix.customized-settings }}" = "default" ]; then
            make e2e-tests LABEL_FILTER="!custom && !joinleave && !resourceplacement"
          elif [ "${{ matrix.customized-settings }}" = "resourceplacement" ]; then
            make e2e-tests LABEL_FILTER="!custom && resourceplacement"
          elif [ "${{ matrix.customized-settings }}" = "joinleave" ]; then
            make e2e-tests LABEL_FILTER="!custom && joinleave"
          else
            make e2e-tests-custom
          fi
        env:
          KUBECONFIG: '/home/runner/.kube/config'
          HUB_SERVER_URL: 'https://172.19.0.2:6443'
          # Temporarily enable the AKS property provider for the E2E tests, in order
          # to verify the property-based scheduling experience.
          #
          # TO-DO (chenyu1): to ensure a vendor-neutral experience, switch to a dummy
          # property provider once the AKS one is split out.
          PROPERTY_PROVIDER: 'azure'
          RESOURCE_SNAPSHOT_CREATION_MINIMUM_INTERVAL: ${{ matrix.resource-snapshot-creation-minimum-interval }}
          RESOURCE_CHANGES_COLLECTION_DURATION: ${{ matrix.resource-changes-collection-duration }}
          CERT_MANAGER_VERSION: ${{ env.CERT_MANAGER_VERSION }}
      
      - name: Collect logs
        if: always()
        # Wait for a bit before log collection; this gives the agent pods some time to shut down
        # gracefully and flush their logs.
        run: |
          sleep 30
          make collect-e2e-logs
        env:
          KUBECONFIG: '/home/runner/.kube/config'
          LOG_DIR: 'logs-${{ matrix.customized-settings }}'
      
      - name: Upload logs
        if: always()
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
        with:
          name: e2e-logs-${{ matrix.customized-settings }}
          path: test/e2e/logs-${{ matrix.customized-settings }}/
          retention-days: 3