integration.yml

# Integration Test Suite
#
# This workflow runs the integration tests. If the workflow is triggered in the merge queue, all integration tests
# are run. If the workflow is triggered in a PR commit, then the files changed in the PR are evaluated to determine
# if any integration tests will run.

name: Integration Test Suite

on:
  workflow_dispatch:
  pull_request:
  merge_group:
    types: [checks_requested]

concurrency:
  # `github.event.number` exists for pull requests, otherwise fall back to SHA for merge queue
  group: ${{ github.workflow }}-${{ github.event.number || github.event.merge_group.head_sha }}
  cancel-in-progress: true

permissions:
  contents: read

env:
  CONTAINER_TOOL: "docker"
  DD_ENV: "ci"
  DD_API_KEY: ${{ secrets.DD_API_KEY }}
  TEST_DATADOG_API_KEY: ${{ secrets.CI_TEST_DATADOG_API_KEY }}
  TEST_APPSIGNAL_PUSH_API_KEY: ${{ secrets.TEST_APPSIGNAL_PUSH_API_KEY }}
  AXIOM_TOKEN: ${{ secrets.AXIOM_TOKEN }}
  RUST_BACKTRACE: full
  VECTOR_LOG: vector=debug
  VERBOSE: true
  CI: true
  PROFILE: debug

jobs:
  changes:
    if: github.event_name == 'merge_group'
    uses: ./.github/workflows/changes.yml
    with:
      source: true
      int_tests: true
      e2e_tests: true
    secrets: inherit

  build-test-runner:
    # Elevated permission required by build-test-runner.yml to push test runner image to GHCR
    permissions:
      contents: read
      packages: write
    needs: changes
    if: ${{
      always() &&
      (github.event_name == 'workflow_dispatch' ||
      (github.event_name == 'merge_group' &&
      needs.changes.result == 'success' &&
      (needs.changes.outputs.dependencies == 'true' ||
      needs.changes.outputs.integration-yml == 'true' ||
      needs.changes.outputs.int-tests-any == 'true' ||
      needs.changes.outputs.e2e-tests-any == 'true')))
      }}
    uses: ./.github/workflows/build-test-runner.yml
    with:
      commit_sha: ${{ github.sha }}

  integration-tests:
    runs-on: ubuntu-24.04-8core
    needs:
      - changes
      - build-test-runner

    if: ${{ always() && !failure() && !cancelled() && needs.build-test-runner.result == 'success' &&
      (github.event_name == 'merge_group' || github.event_name == 'workflow_dispatch') }}
    strategy:
      matrix:
        # TODO: Add "splunk" back once https://github.com/vectordotdev/vector/issues/23474 is fixed.
        # If you modify this list, please also update the `int_tests` job in changes.yml.
        service:
          [
            "amqp",
            "appsignal",
            "axiom",
            "aws",
            "azure",
            "clickhouse",
            "databend",
            "datadog-agent",
            "datadog-logs",
            "datadog-metrics",
            "datadog-traces",
            "dnstap",
            "docker-logs",
            "doris",
            "elasticsearch",
            "eventstoredb",
            "fluent",
            "gcp",
            "greptimedb",
            "http-client",
            "influxdb",
            "kafka",
            "logstash",
            "loki",
            "mqtt",
            "mongodb",
            "nats",
            "nginx",
            "opentelemetry",
            "postgres",
            "prometheus",
            "pulsar",
            "redis",
            "webhdfs"
          ]
    timeout-minutes: 90
    steps:
      - name: Download JSON artifact from changes.yml
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3 # v8.0.0
        if: github.event_name == 'merge_group'
        with:
          name: int_tests_changes

      - name: Determine if test should run
        id: check
        run: |
          if [[ "${{ github.event_name }}" == "workflow_dispatch" || \
                "${{ needs.changes.outputs.dependencies }}" == "true" || \
                "${{ needs.changes.outputs.integration-yml }}" == "true" ]]; then
            echo "should_run=true" >> "$GITHUB_OUTPUT"
          elif [[ "${{ needs.changes.outputs.website_only }}" == "true" ]]; then
            echo "should_run=false" >> "$GITHUB_OUTPUT"
          elif [[ -f int_tests_changes.json ]]; then
            should_run=$(jq -r '."${{ matrix.service }}" // false' int_tests_changes.json)
            echo "should_run=${should_run}" >> "$GITHUB_OUTPUT"
          else
            echo "should_run=false" >> "$GITHUB_OUTPUT"
          fi

      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        if: steps.check.outputs.should_run == 'true'
        with:
          submodules: "recursive"

      - uses: ./.github/actions/setup
        if: steps.check.outputs.should_run == 'true'
        with:
          vdev: true
          mold: false
          cargo-cache: false
          datadog-ci: true

      - name: Pull test runner image
        if: steps.check.outputs.should_run == 'true'
        uses: ./.github/actions/pull-test-runner
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}
          commit_sha: ${{ github.sha }}

      - name: Run Integration Tests for ${{ matrix.service }}
        if: steps.check.outputs.should_run == 'true'
        uses: nick-fields/retry@ce71cc2ab81d554ebbe88c79ab5975992d79ba08 # v3.0.2
        with:
          timeout_minutes: 30
          max_attempts: 3
          command: |
            bash scripts/run-integration-test.sh int ${{ matrix.service }}

  e2e-tests:
    runs-on: ubuntu-24.04-8core
    needs:
      - changes
      - build-test-runner
    if: ${{ always() && !failure() && !cancelled() && needs.build-test-runner.result == 'success' &&
      (github.event_name == 'merge_group' || github.event_name == 'workflow_dispatch') }}
    strategy:
      matrix:
        service: ["datadog-logs", "datadog-metrics", "opentelemetry-logs", "opentelemetry-metrics"]
    timeout-minutes: 90

    steps:
      - name: Download JSON artifact from changes.yml
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3 # v8.0.0
        if: github.event_name == 'merge_group'
        with:
          name: e2e_tests_changes

      - name: Determine if test should run
        id: check
        run: |
          if [[ "${{ github.event_name }}" == "workflow_dispatch" || \
                "${{ needs.changes.outputs.dependencies }}" == "true" || \
                "${{ needs.changes.outputs.integration-yml }}" == "true" ]]; then
            echo "should_run=true" >> "$GITHUB_OUTPUT"
          elif [[ "${{ needs.changes.outputs.website_only }}" == "true" ]]; then
            echo "should_run=false" >> "$GITHUB_OUTPUT"
          elif [[ -f e2e_tests_changes.json ]]; then
            should_run=$(jq -r '."${{ matrix.service }}" // false' e2e_tests_changes.json)
            echo "should_run=${should_run}" >> "$GITHUB_OUTPUT"
          else
            echo "should_run=false" >> "$GITHUB_OUTPUT"
          fi

      - uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2
        if: steps.check.outputs.should_run == 'true'
        with:
          submodules: "recursive"

      - uses: ./.github/actions/setup
        if: steps.check.outputs.should_run == 'true'
        with:
          vdev: true
          mold: false
          cargo-cache: false
          datadog-ci: true

      - name: Pull test runner image
        if: steps.check.outputs.should_run == 'true'
        uses: ./.github/actions/pull-test-runner
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}
          commit_sha: ${{ github.sha }}

      - name: Run E2E Tests for ${{ matrix.service }}
        if: steps.check.outputs.should_run == 'true'
        uses: nick-fields/retry@ce71cc2ab81d554ebbe88c79ab5975992d79ba08 # v3.0.2
        with:
          timeout_minutes: 30
          max_attempts: 3
          command: |
            bash scripts/run-integration-test.sh e2e ${{ matrix.service }}

  integration-test-suite:
    name: Integration Test Suite
    runs-on: ubuntu-24.04
    if: always()
    needs:
      - changes
      - build-test-runner
      - integration-tests
      - e2e-tests
    steps:
      - run: |
          if [[ "${{ contains(needs.*.result, 'failure') || contains(needs.*.result, 'cancelled') }}" == "true" ]]; then
            echo "One or more jobs failed or were cancelled"
            exit 1
          else
            echo "All jobs completed successfully"
          fi