my-rook-config/.github/workflows/multus.yaml

name: Multus integration tests
on:
  push:
    tags:
      - v*
    branches:
      - master
      - release-*
  pull_request:
    paths:
      - cmd/rook/userfacing/**
      - pkg/daemon/multus/**
      - .github/workflows/multus.yaml
      - tests/scripts/multus/**

defaults:
  run:
    # reference: https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#using-a-specific-shell
    shell: bash --noprofile --norc -eo pipefail -x {0}

# cancel the in-progress workflow when PR is refreshed.
concurrency:
  group: ${{ github.workflow }}-${{ github.event_name == 'pull_request' && github.head_ref || github.sha }}
  cancel-in-progress: true

permissions:
  contents: read

jobs:
  test-validation-tool:
    runs-on: ubuntu-latest
    strategy:
      fail-fast: true
    if: "!contains(github.event.pull_request.labels.*.name, 'skip-ci')"
    steps:
      - name: checkout
        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          fetch-depth: 0

      - name: Set up Go version
        uses: actions/setup-go@b7ad1dad31e06c5925ef5d2fc7ad053ef454303e # v7.0.0
        with:
          go-version: "1.26"

      - name: Create KinD Cluster
        uses: helm/kind-action@ef37e7f390d99f746eb8b610417061a60e82a6cc # v1.14.0
        with:
          config: tests/config/multus/kind-config.yaml
          cluster_name: kind

      - name: consider (pre-job) debugging
        uses: ./.github/workflows/tmate_debug
        with:
          use-tmate: ${{ secrets.USE_TMATE }}
          debug-ci: ${{ contains(github.event.pull_request.labels.*.name, 'debug-ci') }}

      - name: Setup multus
        run: ./tests/scripts/multus/setup-multus.sh

      - name: Set up multus prerequisite host routing
        run: |
          kubectl create -f tests/scripts/multus/host-cfg-ds.yaml
          kubectl rollout status daemonset/host-net-config --timeout=120s

      - name: Install public and cluster NADs in default namespace
        run: kubectl create -f tests/scripts/multus/default-public-cluster-nads.yaml

      - name: Quickly build Rook binary
        id: build
        run: go build -tags=ceph_preview -o rook cmd/rook/*.go

      - name: Run CLI validation test
        run: ./tests/scripts/multus/test-110-cli.sh

      - name: Label kind nodes for stretch cluster tests
        id: label
        # tests that require labels are independent from previous tests
        if: steps.build.outcome == 'success' && !cancelled()
        run: ./tests/scripts/multus/test-200-stretch-label-nodes.sh
        # nodes are purposefully not tainted yet for overlap test

      - name: Run stretch cluster overlap test
        id: overlap
        # independent from other tests as long as nodes are labeled
        if: steps.label.outcome == 'success' && !cancelled()
        run: ./tests/scripts/multus/test-210-stretch-overlap.sh

      - name: Run cleanup test
        # cleanup relies on overlap test to be successful
        if: steps.overlap.outcome == 'success' && !cancelled()
        run: ./tests/scripts/multus/test-211-stretch-cleanup.sh

      - name: Taint kind nodes for remaining stretch cluster tests
        id: taint
        # tests that require taints+labels are independent from previous tests
        if: steps.label.outcome == 'success' && !cancelled()
        run: ./tests/scripts/multus/test-200-stretch-taint-nodes.sh

      - name: Run stretch cluster test with public and cluster networks
        # independent from other tests as long as nodes are tainted and labeled
        if: steps.taint.outcome == 'success' && !cancelled()
        run: ./tests/scripts/multus/test-220-stretch-pub-and-cluster.sh

      - name: Run stretch cluster test with public network only
        # independent from other tests as long as nodes are tainted and labeled
        if: steps.taint.outcome == 'success' && !cancelled()
        run: ./tests/scripts/multus/test-230-stretch-pub-only.sh

      - name: Run stretch cluster test with cluster network only
        # independent from other tests as long as nodes are tainted and labeled
        if: steps.taint.outcome == 'success' && !cancelled()
        run: ./tests/scripts/multus/test-240-stretch-cluster-only.sh

      - name: dump cluster state on failure
        if: failure()
        run: |
          kubectl get nodes -o wide || true
          kubectl get pods -A -o wide || true
          kubectl -n kube-system describe daemonset kube-multus-ds install-cni-plugins whereabouts || true
          kubectl get events -A --sort-by=.lastTimestamp | tail -60 || true
          kubectl -n kube-system logs -l name=multus --all-containers --tail=30 || true

      - name: consider (post-job) debugging
        uses: ./.github/workflows/upterm_debug
        with:
          debug-ci: ${{ contains(github.event.pull_request.labels.*.name, 'debug-ci') }}