weave-trace-tests.yaml

name: weave-trace-tests

on:
  push:
  workflow_dispatch:

permissions:
  contents: read
  id-token: write

jobs:
  check-which-tests-to-run:
    uses: ./.github/workflows/check-which-tests-to-run.yaml

  # This duplicates core's Build and Test App / Weave Trace Tests job while
  # trace-server still lives in core. Simplify this once trace-server moves into
  # weave or the core workflow can be reused directly.
  weave-trace-tests:
    name: Weave Trace Tests
    needs:
      - check-which-tests-to-run
    if: needs.check-which-tests-to-run.outputs.trace_server_tests == 'true' || github.ref == 'refs/heads/master' || github.event_name == 'workflow_dispatch'
    runs-on: ubuntu-latest-8-cores
    timeout-minutes: 45

    env:
      CORE_DIR: core
      COMPOSE_FILE: .github/compose/weave-trace-tests.yaml
      COMPOSE_PROJECT_NAME: weave-trace-tests

    steps:
      - name: Create GitHub App token
        id: app-token
        uses: actions/create-github-app-token@bcd2ba49218906704ab6c1aa796996da409d3eb1 # v3
        with:
          app-id: ${{ vars.WANDBOT_3000_APP_ID }}
          private-key: ${{ secrets.WANDBOT_3000_PRIVATE_KEY }}
          owner: wandb
          repositories: core
          permission-contents: read

      - name: Checkout core
        uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
        with:
          repository: wandb/core
          ref: master
          path: ${{ env.CORE_DIR }}
          token: ${{ steps.app-token.outputs.token }}
          persist-credentials: false

      - name: Remove core Weave submodule placeholder
        run: rm -rf "${CORE_DIR}/services/weave-python/weave-public"

      - name: Checkout current weave commit into core
        uses: actions/checkout@df4cb1c069e1874edd31b4311f1884172cec0e10 # v6
        with:
          path: ${{ env.CORE_DIR }}/services/weave-python/weave-public
          persist-credentials: false

      - name: Verify overlaid Weave commit
        run: |
          set -euo pipefail

          actual_sha="$(git -C "${CORE_DIR}/services/weave-python/weave-public" rev-parse HEAD)"
          echo "core_sha=$(git -C "${CORE_DIR}" rev-parse HEAD)"
          echo "weave_sha=${actual_sha}"
          test "${actual_sha}" = "${GITHUB_SHA}"

      - name: Authenticate to Google Cloud
        uses: google-github-actions/auth@7c6bc770dae815cd3e89ee6cdf493a5fab2cc093 # v3
        with:
          workload_identity_provider: ${{ secrets.WORKLOAD_IDENTITY_PROVIDER }}
          service_account: ${{ secrets.WORKLOAD_IDENTITY_SERVICE_ACCOUNT }}

      - name: Set up gcloud
        uses: google-github-actions/setup-gcloud@aa5489c8933f4cc7a4f7d45035b3b1440c9c10db # v3

      - name: Configure Docker auth
        run: gcloud auth configure-docker us-central1-docker.pkg.dev --quiet

      - name: Start service containers
        working-directory: ${{ env.CORE_DIR }}
        run: docker compose up --detach

      - name: Install uv
        uses: astral-sh/setup-uv@37802adc94f370d6bfd71619e3f0bf239e1f3b78 # v7
        with:
          enable-cache: true
          cache-dependency-glob: core/services/weave-trace/uv.lock

      - name: Install dependencies
        working-directory: ${{ env.CORE_DIR }}/services/weave-trace
        run: uv sync --group dev --group test --frozen

      - name: Pip freeze
        working-directory: ${{ env.CORE_DIR }}/services/weave-trace
        run: uv pip freeze

      - name: Wait for ClickHouse
        run: |
          set -euo pipefail
          for _i in {1..60}; do
            if curl --silent --fail --show-error http://localhost:8123/ping; then
              echo "ClickHouse is ready"
              exit 0
            fi
            sleep 2
          done
          echo "::error::ClickHouse did not become ready in time"
          exit 1

      - name: Wait for Gorilla
        run: |
          set -euo pipefail
          for _i in {1..120}; do
            if (curl --silent --fail --show-error http://localhost:8080/ready >/dev/null \
                || curl --silent --fail --show-error http://localhost:8080/healthz >/dev/null) \
              && curl --silent --fail --show-error http://localhost:9015/health >/dev/null; then
              echo "Gorilla and fixture service are ready"
              exit 0
            fi
            sleep 2
          done
          echo "::error::Gorilla did not become ready in time"
          exit 1

      - name: Wait for Bufstream
        run: |
          set -euo pipefail
          for _i in {1..60}; do
            if timeout 5 bash -c "</dev/tcp/localhost/9092"; then
              echo "Bufstream is ready"
              exit 0
            fi
            sleep 1
          done
          echo "::error::Bufstream did not become ready in time"
          exit 1

      - name: Export API schema
        working-directory: ${{ env.CORE_DIR }}/services/weave-trace
        run: |
          export PYTHONPATH="${GITHUB_WORKSPACE}/${CORE_DIR}/services/weave-python/weave-public${PYTHONPATH:+:${PYTHONPATH}}"
          make export-api-schema

      - name: Run Weave Trace tests
        working-directory: ${{ env.CORE_DIR }}/services/weave-trace
        env:
          DD_ENV: ci
          WB_SERVER_HOST: http://localhost
          WF_CLICKHOUSE_HOST: localhost
          WEAVE_TRACE_CLICKHOUSE_USE_ASYNC_INSERT: "false"
          BUFSTREAM_KAFKA_HOST: localhost
          KAFKA_BROKER_HOST: localhost
          KAFKA_BROKER_PORT: "9092"
        run: |
          export PYTHONPATH="${GITHUB_WORKSPACE}/${CORE_DIR}/services/weave-python/weave-public${PYTHONPATH:+:${PYTHONPATH}}"
          make test-cov

      - name: Dump service logs
        if: failure()
        working-directory: ${{ env.CORE_DIR }}
        run: |
          docker compose ps || true
          for service in wandb-local clickhouse bufstream; do
            echo "::group::${service} logs"
            docker compose logs --no-color "${service}" || true
            echo "::endgroup::"
          done

      - name: Stop service containers
        if: always()
        working-directory: ${{ env.CORE_DIR }}
        run: docker compose down --volumes --remove-orphans || true