unit-tests-elastic.yml

name: Elastic Module Tests

on:
  push:
    branches: [ main, develop ]
    paths:
      - 'flagscale/runner/elastic/**'
      - 'tests/unit_tests/runner/elastic/**'
      - '.github/workflows/unit-tests-elastic.yml'
  pull_request:
    branches: [ main, develop ]
    paths:
      - 'flagscale/runner/elastic/**'
      - 'tests/unit_tests/runner/elastic/**'
      - '.github/workflows/unit-tests-elastic.yml'

jobs:
  lint-and-test:
    runs-on: [self-hosted, Linux, X64, nvidia-0, gpus-8]
    container:
      image: flagscale/flagscale:dev-megatron
      ports:
        - 80
      volumes:
        - /home/flagscale_cicd/flask/static:/workspace/report
        - /home/flagscale_cicd/flask/config:/workspace/config
        - /home/flagscale_cicd/docker/docker_build/docker_data:/home/gitlab-runner/data
        - /home/flagscale_cicd/docker/docker_build/docker_tokenizers:/home/gitlab-runner/tokenizers
      options: --gpus all --shm-size=500g --hostname flagscale_cicd --user root --ulimit nofile=65535:65535
    
    steps:
      - name: Checkout Code
        uses: actions/checkout@v4
        with:
          repository: ${{ github.event.pull_request.head.repo.full_name || github.repository }}
          ref: ${{ github.event.pull_request.head.ref || github.ref }}
          ssh-strict: true
          ssh-user: git
          persist-credentials: true
          clean: true
          sparse-checkout-cone-mode: true
          fetch-tags: false
          show-progress: true
          lfs: false
          submodules: false
          set-safe-directory: true

      - name: Setup Environment
        run: |
          echo "USER: $USER"
          echo "UID: $(id -u)"
          echo "GID: $(id -g)"
          echo "Home: $HOME"
          whoami
          git config --global --add safe.directory /__w/FlagScale/FlagScale || true

      - name: Install Dependencies
        run: |
          source /root/miniconda3/etc/profile.d/conda.sh
          conda activate flagscale-train
          pip install pylint pytest pytest-cov
          python tools/patch/unpatch.py --backend Megatron-LM || true
          export PYTHONPATH=./third_party/Megatron-LM:$PYTHONPATH
          export NVTE_FLASH_ATTN=0
          export NVTE_FUSED_ATTN=0
          ulimit -n 65535

      - name: Pylint Check - Elastic Module
        run: |
          source /root/miniconda3/etc/profile.d/conda.sh
          conda activate flagscale-train
          export PYTHONPATH=./third_party/Megatron-LM:$PYTHONPATH
          
          echo "Running pylint on elastic module..."
          pylint flagscale/runner/elastic/ --output-format=text --reports=yes --exit-zero > pylint_report.txt
          
          # Display pylint results
          cat pylint_report.txt
          
          # Extract pylint score
          PYLINT_SCORE=$(grep "Your code has been rated at" pylint_report.txt | sed 's/.*rated at \([0-9.]*\).*/\1/' || echo "0")
          echo "Pylint Score: $PYLINT_SCORE"
          
          # Fail if score is below 8.0
          if (( $(echo "$PYLINT_SCORE < 8.0" | bc -l) )); then
            echo "Pylint score $PYLINT_SCORE is below required threshold of 8.0"
            exit 1
          fi

      - name: Run Elastic Unit Tests
        run: |
          source /root/miniconda3/etc/profile.d/conda.sh
          conda activate flagscale-train
          export PYTHONPATH=./third_party/Megatron-LM:$PYTHONPATH
          export NVTE_FLASH_ATTN=0
          export NVTE_FUSED_ATTN=0
          ulimit -n 65535
          
          echo "Running elastic module unit tests..."
          pytest tests/unit_tests/runner/elastic/ \
            --cov=flagscale/runner/elastic \
            --cov-report=xml:coverage_elastic.xml \
            --cov-report=html:coverage_elastic_html \
            --cov-report=term \
            -v \
            --tb=short \
            -x

      - name: Coverage Report
        run: |
          source /root/miniconda3/etc/profile.d/conda.sh
          conda activate flagscale-train
          
          echo "Coverage Summary for Elastic Module:"
          if [ -f coverage_elastic.xml ]; then
            echo "Coverage XML report generated successfully"
            python -c "
import xml.etree.ElementTree as ET
try:
    tree = ET.parse('coverage_elastic.xml')
    root = tree.getroot()
    coverage = root.attrib.get('line-rate', '0')
    percentage = float(coverage) * 100
    print(f'Line Coverage: {percentage:.1f}%')
    if percentage < 80:
        print(f'Warning: Coverage {percentage:.1f}% is below recommended 80%')
        exit(1)
except Exception as e:
    print(f'Error parsing coverage: {e}')
    exit(1)
"
          else
            echo "Coverage XML report not found"
            exit 1
          fi

      - name: Archive Test Results
        uses: actions/upload-artifact@v3
        if: always()
        with:
          name: elastic-test-results
          path: |
            pylint_report.txt
            coverage_elastic.xml
            coverage_elastic_html/
          retention-days: 30

      - name: Summary
        if: always()
        run: |
          echo "=== Elastic Module CI/CD Summary ==="
          echo "[INFO] Pylint check completed"
          echo "[INFO] Unit tests executed" 
          echo "[INFO] Coverage report generated"
          echo "=== Test artifacts uploaded to GitHub Actions ==="