post-workflow-actions.yml

name: Post Workflow Actions
run-name: Post Workflow Actions for ${{ github.event_name == 'workflow_dispatch' && format('run {0}', inputs.run_id) || github.event.workflow_run.name }}

# Triggered when a monitored workflow completes.
# NOTE: workflow_run.workflows does not support wildcards —
# workflow names must be listed explicitly here.
# This workflow is NOT listed, so it will not trigger itself (no recursion).
# Listing workflows that run in PR and Post Commit to the list
on:
  workflow_run:
    workflows:
      - "AI Smoke Tests (Local Embeddings)"
      - "Airflow Plugin"
      - "build & test"
      - "Check Datahub Jars"
      - "check python dependencies"
      - "CI Cache Warmup"
      - "Dagster Plugin"
      - "DataHub Actions"
      - "DataHub Agent Context"
      - "Docker Build, Scan, Test"
      - "documentation"
      - "Frontend Preview"
      - "GX Plugin"
      - "ingestion smoke"
      - "lint"
      - "Metadata Ingestion"
      - "metadata model generate"
      - "Metadata Models Custom CI"
      - "metadata-io"
      - "Nightly Docker Test"
      - "OAuth Smoke Test"
      - "PR Title Check"
      - "Prefect Plugin"
      - "Publish Datahub Java Jars (Client, Spark Lineage, Protobuf, Auth API)"
      - "Python Build"
      - "Python Tests for github scripts"
      - "Quickstart Test"
      - "Release Tests"
      - "Release Validation"
      - "spark smoke test"
      - "Vercel Preview Cleanup"
      - "Verify Quickstart Compose"
    types:
      - completed

  # Allows manual triggering for ad-hoc testing against any existing workflow run.
  # publish defaults to false so artifacts are not forwarded to downstream by accident.
  workflow_dispatch:
    inputs:
      run_id:
        description: "Workflow run ID to collect metrics for"
        required: true
        type: string
      attempt:
        description: "Run attempt number"
        required: false
        type: string
        default: "1"
      publish:
        description: "Publish metrics artifacts downstream"
        required: false
        type: boolean
        default: false
      notify_slack:
        description: "Send Slack failure notification (for testing; production
          notifications fire automatically)"
        required: false
        type: boolean
        default: false

permissions:
  actions: read
  contents: read

jobs:
  collect-metrics:
    # For push events: only collect metrics for default/release/hotfix branches.
    # All other event types (pull_request, schedule, release, etc.) and workflow_dispatch pass through.
    if: >-
      github.event_name == 'workflow_dispatch' ||
      github.event.workflow_run.event != 'push' ||
      github.event.workflow_run.head_branch == github.event.repository.default_branch ||
      startsWith(github.event.workflow_run.head_branch, 'releases/') ||
      startsWith(github.event.workflow_run.head_branch, 'hotfixes/')
    runs-on: ubuntu-22.04
    steps:
      - name: Checkout repository
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd # v6.0.2

      - name: Set up Python
        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405 # v6
        with:
          python-version: "3.12"

      - name: Install uv
        uses: astral-sh/setup-uv@37802adc94f370d6bfd71619e3f0bf239e1f3b78 # v7.6.0
        with:
          enable-cache: true

      - name: Install dependencies
        run: uv pip install requests==2.32.5 posthog==7.9.4 --system

      - name: Collect metrics
        id: collect-metrics
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
          # Fall back to workflow_run event values when not triggered manually.
          RUN_ID: ${{ inputs.run_id || github.event.workflow_run.id }}
          ATTEMPT: ${{ inputs.attempt || github.event.workflow_run.run_attempt }}
          # Always publish for automated workflow_run triggers; honour the explicit input for workflow_dispatch.
          PUBLISH: ${{ github.event_name == 'workflow_run' || inputs.publish == true }}
        run: |
          ARTIFACT_NAME="workflow-metrics-${RUN_ID}-attempt-${ATTEMPT}"
          echo "artifact_name=${ARTIFACT_NAME}" >> "$GITHUB_OUTPUT"
          echo "publish=${PUBLISH}" >> "$GITHUB_OUTPUT"
          cd .github/scripts
          python collect_workflow_metrics.py \
            --run-id "${RUN_ID}" \
            --attempt "${ATTEMPT}" \
            --repo "${{ github.repository }}" \
            --output "${ARTIFACT_NAME}.json"
          # Print workflow summary
          jq -r '
            "## \(.workflow.name // "Workflow"): \(.workflow.conclusion // "unknown")",
            "Duration: \(.workflow.duration_seconds // 0)s | Jobs: \(.jobs | length)",
            "",
            "| Job | Conclusion | Steps | Duration |",
            "|---|---|---|---|",
            (.jobs[] | "| \(.full_name // .name) | \(.conclusion // "unknown") | \(.steps | length) | \(.duration_seconds // 0)s |")
          ' "${ARTIFACT_NAME}.json" | tee -a "${GITHUB_STEP_SUMMARY}"

      - name: Upload metrics artifact
        uses: actions/upload-artifact@bbbca2ddaa5d8feaa63e36b76fdaad77386f024f # v7.0.0
        with:
          name: ${{ steps.collect-metrics.outputs.artifact_name }}
          path: .github/scripts/${{ steps.collect-metrics.outputs.artifact_name }}.json
          if-no-files-found: error
          retention-days: 7
      - name: Publish metrics to PostHog
        if: steps.collect-metrics.outputs.publish == 'true'
        env:
          POSTHOG_API_KEY: ${{ secrets.POSTHOG_API_KEY }}
        run: |
          cd .github/scripts
          python publish_to_posthog.py \
            --input "${{ steps.collect-metrics.outputs.artifact_name }}.json"
      # TODO(devashish.chandra): Save to S3?

  retry-on-failure:
    name: Retry on First Failure
    runs-on: ubuntu-22.04
    # Only retrigger workflows known to have flaky tests, on first failure only,
    # and only when triggered by workflow_run (not workflow_dispatch).
    if: >-
      github.event_name == 'workflow_run' &&
      (
        github.event.workflow_run.name == 'Docker Build, Scan, Test' ||
        github.event.workflow_run.name == 'metadata-io' ||
        github.event.workflow_run.name == 'Nightly Docker Test'
      ) &&
      github.event.workflow_run.conclusion == 'failure' &&
      github.event.workflow_run.run_attempt == 1
    permissions:
      actions: write
    steps:
      - name: Rerun failed jobs
        continue-on-error: true
        env:
          GH_TOKEN: ${{ github.token }}
          RUN_ID: ${{ github.event.workflow_run.id }}
          REPOSITORY: ${{ github.repository }}
        run: |
          gh api \
            --method POST \
            -H "Accept: application/vnd.github+json" \
            -H "X-GitHub-Api-Version: 2022-11-28" \
            "/repos/${REPOSITORY}/actions/runs/${RUN_ID}/rerun-failed-jobs"

  # Notify Slack on CI failure on protected branches.
  # Fires when:
  #   - workflow_dispatch with notify_slack=true (manual testing)
  #   - workflow_run that failed or timed out,
  #       - triggered by schedule/release
  #       - or a push to a protected branch (default, release/*, hotfix/*)
  notify-slack-failure:
    name: Notify Slack on CI failure
    if: >-
      (github.event_name == 'workflow_dispatch' && inputs.notify_slack == true)
      || (github.event_name == 'workflow_run'
          && contains(fromJSON('["failure","timed_out"]'), github.event.workflow_run.conclusion)
          && (contains(fromJSON('["schedule","release"]'), github.event.workflow_run.event)
              || (github.event.workflow_run.event == 'push'
                  && (github.event.workflow_run.head_branch == github.event.repository.default_branch
                      || startsWith(github.event.workflow_run.head_branch, 'releases/')
                      || startsWith(github.event.workflow_run.head_branch, 'hotfixes/')))))
    uses: ./.github/workflows/notify-slack-status.yml
    secrets: inherit
    with:
      run_id: ${{ inputs.run_id || github.event.workflow_run.id }}
      attempt: ${{ inputs.attempt || github.event.workflow_run.run_attempt }}
      thread_ts_file: ${{ github.event.workflow_run.event != 'schedule' && '.slack-thread-ts.json' || '' }}
      channel: ${{ vars.BUILD_STATUS_NOTIFICATION_SLACK_CHANNEL }}

  # Notify Slack when the Docker build succeeds on a push or release event.
  # Posts a standalone message (not threaded) so it's visible as a positive signal
  # separate from any failure thread on the same commit.
  notify-slack-docker-success:
    name: Notify Slack on Docker build success
    if: >-
      github.event_name == 'workflow_run'
      && github.event.workflow_run.conclusion == 'success'
      && github.event.workflow_run.name == 'Docker Build, Scan, Test'
      && contains(fromJSON('["push","release"]'), github.event.workflow_run.event)
    uses: ./.github/workflows/notify-slack-status.yml
    secrets: inherit
    with:
      run_id: ${{ github.event.workflow_run.id }}
      attempt: ${{ github.event.workflow_run.run_attempt }}
      conclusion: success
      channel: ${{ vars.BUILD_STATUS_NOTIFICATION_SLACK_CHANNEL }}