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)"
      - "DataHub Actions"
      - "DataHub Agent Context"
      - "Airflow Plugin"
      - "build & test"
      - "Check Datahub Jars"
      - "check python dependencies"
      - "Dagster Plugin"
      - "Docker Build, Scan, Test"
      - "ingestion smoke"
      - "documentation"
      - "GX Plugin"
      - "lint"
      - "Metadata Ingestion"
      - "metadata-io"
      - "metadata model generate"
      - "Metadata Models Custom CI"
      - "Prefect Plugin"
      - "PR Title Check"
      - "Publish Datahub Java Jars (Client, Spark Lineage, Protobuf, Auth API)"
      - "Python Build"
      - "Quickstart Test"
      - "Frontend Preview"
      - "spark smoke test"
      - "Verify Quickstart Compose"
      - "Nightly Docker Test"
      - "Release Tests"
      - "Release Validation"
    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

permissions:
  actions: read
  contents: read

jobs:
  collect-metrics:
    # For workflow_run: only collect metrics for PRs and pushes to the default/release/hotfix
    # branches — skip feature branches, dependabot bumps, etc.
    # workflow_dispatch is always allowed for ad-hoc testing.
    if: >-
      github.event_name == 'workflow_dispatch' ||
      github.event.workflow_run.event == 'pull_request' ||
      (github.event.workflow_run.event == 'push' &&
        (github.event.workflow_run.head_branch == github.event.repository.default_branch ||
         startsWith(github.event.workflow_run.head_branch, 'release/') ||
         startsWith(github.event.workflow_run.head_branch, 'hotfix/')))
    runs-on: ubuntu-22.04
    steps:
      - name: Checkout repository
        uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5 # v4

      - name: Set up Python
        uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5
        with:
          python-version: "3.12"

      - name: Install uv
        uses: astral-sh/setup-uv@d4b2f3b6ecc6e67c4457f6d3e41ec42d3d0fcb86 # v5
        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@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4
        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"