Try a different approach to parallel tests #6605
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| name: Install and Test Firedrake | |
| on: | |
| # Push to master or PR | |
| push: | |
| branches: | |
| - master | |
| pull_request: | |
| schedule: | |
| - cron: '0 0 * * 0' | |
| - cron: '0 0 1 * *' # Monthly release | |
| concurrency: | |
| # Cancels jobs running if new commits are pushed | |
| group: > | |
| ${{ github.workflow }}- | |
| ${{ github.event.pull_request.number || github.ref }} | |
| cancel-in-progress: true | |
| env: | |
| RELEASE_TAG: latest | |
| jobs: | |
| test: | |
| name: "Run Firedrake tests (Linux)" | |
| runs-on: [self-hosted, Linux] | |
| container: | |
| image: firedrakeproject/firedrake-env:latest | |
| strategy: | |
| fail-fast: false | |
| matrix: | |
| include: | |
| - scalar-type: real | |
| complex: "" | |
| petsc_arch: default | |
| - scalar-type: complex | |
| complex: --complex | |
| petsc_arch: complex | |
| env: | |
| # PETSC_DIR and MPICH_DIR are set inside the docker image | |
| FIREDRAKE_CI_TESTS: 1 | |
| PYOP2_CI_TESTS: 1 | |
| PETSC_ARCH: ${{ matrix.petsc_arch }} | |
| OMP_NUM_THREADS: 1 | |
| OPENBLAS_NUM_THREADS: 1 | |
| COMPLEX: ${{ matrix.complex }} | |
| RDMAV_FORK_SAFE: 1 | |
| outputs: | |
| scalar-type: ${{ matrix.scalar-type }} | |
| steps: | |
| - uses: actions/checkout@v4 | |
| - name: Cleanup | |
| if: ${{ always() }} | |
| run: | | |
| cd .. | |
| rm -rf firedrake_venv | |
| - name: Build Firedrake | |
| run: | | |
| cd .. | |
| # Linting should ignore unquoted shell variable $COMPLEX | |
| # shellcheck disable=SC2086 | |
| ./firedrake/scripts/firedrake-install \ | |
| $COMPLEX \ | |
| --honour-petsc-dir \ | |
| --mpicc="$MPICH_DIR"/mpicc \ | |
| --mpicxx="$MPICH_DIR"/mpicxx \ | |
| --mpif90="$MPICH_DIR"/mpif90 \ | |
| --mpiexec="$MPICH_DIR"/mpiexec \ | |
| --mpihome="$MPICH_DIR"/.. \ | |
| --venv-name firedrake_venv \ | |
| --no-package-manager \ | |
| --disable-ssh \ | |
| --documentation-dependencies \ | |
| --torch \ | |
| --jax \ | |
| --netgen \ | |
| --slepc \ | |
| --install thetis \ | |
| --install gusto \ | |
| --install icepack \ | |
| --install irksome \ | |
| --install femlium \ | |
| --install fascd \ | |
| --install defcon \ | |
| --install gadopt \ | |
| --install asQ \ | |
| || (cat firedrake-install.log && /bin/false) | |
| - name: Install test dependencies | |
| id: build | |
| run: | | |
| sudo apt update | |
| sudo apt -y install parallel | |
| . ../firedrake_venv/bin/activate | |
| python "$(which firedrake-clean)" | |
| python -m pip install pytest-timeout ipympl pytest-split | |
| python -m pip list | |
| - name: Run tests (nprocs = 1) | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| : # use --quote to stop parallel from parsing the pytest arguments | |
| parallel --line-buffer --tag --quote \ | |
| pytest -v --splits 12 --group {#} --splitting-algorithm least_duration \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake1_{#}.xml \ | |
| -m "parallel[1] or not parallel" tests/firedrake ::: $(seq 12) | |
| - name: Run tests (nprocs = 2) | |
| # Run even if earlier tests failed | |
| if: ${{ success() || steps.build.conclusion == 'success' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| parallel --line-buffer --tag --quote \ | |
| mpiexec -n 2 pytest -v --splits 6 --group {#} --splitting-algorithm least_duration \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake2_{#}.xml \ | |
| -m parallel[2] tests/firedrake ::: $(seq 6) | |
| - name: Run tests (nprocs = 3) | |
| if: ${{ success() || steps.build.conclusion == 'success' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| parallel --line-buffer --tag --quote \ | |
| mpiexec -n 3 pytest -v --splits 4 --group {#} --splitting-algorithm least_duration \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake3_{#}.xml \ | |
| -m parallel[3] tests/firedrake ::: $(seq 4) | |
| - name: Run tests (nprocs = 4) | |
| if: ${{ success() || steps.build.conclusion == 'success' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| parallel --line-buffer --tag --quote \ | |
| mpiexec -n 4 pytest -v --splits 3 --group {#} --splitting-algorithm least_duration \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake4_{#}.xml \ | |
| -m parallel[4] tests/firedrake ::: $(seq 3) | |
| - name: Run tests (nprocs = 5) | |
| if: ${{ success() || steps.build.conclusion == 'success' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| parallel --line-buffer --tag --quote \ | |
| mpiexec -n 5 pytest -v --splits 2 --group {#} --splitting-algorithm least_duration \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake5_{#}.xml \ | |
| -m parallel[5] tests/firedrake ::: $(seq 2) | |
| - name: Run tests (nprocs = 6) | |
| if: ${{ success() || steps.build.conclusion == 'success' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| parallel --line-buffer --tag --quote \ | |
| mpiexec -n 6 pytest -v --splits 2 --group {#} --splitting-algorithm least_duration \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake6_{#}.xml \ | |
| -m parallel[6] tests/firedrake ::: $(seq 2) | |
| - name: Run tests (nprocs = 7) | |
| if: ${{ success() || steps.build.conclusion == 'success' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| mpiexec -n 7 pytest -v \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake7.xml \ | |
| -m parallel[7] tests/firedrake | |
| - name: Run tests (nprocs = 8) | |
| if: ${{ success() || steps.build.conclusion == 'success' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| mpiexec -n 8 pytest -v \ | |
| --timeout=1800 --timeout-method=thread -o faulthandler_timeout=1860 \ | |
| --junit-xml=firedrake8.xml \ | |
| -m parallel[8] tests/firedrake | |
| - name: Publish Test Report | |
| uses: mikepenz/[email protected] | |
| if: ${{ always() && ( github.ref != 'refs/heads/master') }} | |
| with: | |
| report_paths: '/__w/firedrake/firedrake/firedrake*.xml' | |
| comment: true | |
| check_name: "Firedrake ${{ matrix.scalar-type }}" | |
| updateComment: true | |
| flaky_summary: true | |
| - name: Test pyadjoint | |
| if: ${{ matrix.scalar-type == 'real' }} | |
| run: | | |
| . ../firedrake_venv/bin/activate | |
| cd ../firedrake_venv/src/pyadjoint | |
| python -m pytest \ | |
| --strict-markers \ | |
| --durations=200 \ | |
| --timeout=600 \ | |
| --timeout-method=thread \ | |
| -o faulthandler_timeout=660 \ | |
| -n 12 --dist worksteal \ | |
| -sv tests/firedrake_adjoint | |
| timeout-minutes: 30 | |
| - name: Cleanup | |
| # Belt and braces: clean up before and after the run. | |
| if: ${{ always() }} | |
| run: | | |
| cd .. | |
| rm -rf firedrake_venv | |
| docker_tag: | |
| name: "Set the Docker release tag" | |
| runs-on: [self-hosted, Linux] | |
| if: ${{ github.ref == 'refs/heads/master' }} | |
| steps: | |
| - name: Set release tag | |
| # Set a release tag if triggered by monthly CRON job | |
| if: github.event.schedule == '0 0 1 * *' | |
| run: | | |
| DATE_TAG="$(date +%Y-%m)" | |
| echo "RELEASE_TAG=$DATE_TAG" >> "$GITHUB_ENV" | |
| - name: Print release tag being used | |
| run: | | |
| echo The release tag is "$RELEASE_TAG" | |
| outputs: | |
| tag: ${{ env.RELEASE_TAG }} | |
| docker: | |
| name: "Build Docker containers" | |
| # Only run on master, but always generate firedrake-env image, | |
| # even if build fails (see docker.yml) | |
| if: ${{ (github.ref == 'refs/heads/master') && always() }} | |
| needs: [test, docker_tag] | |
| uses: ./.github/workflows/docker.yml | |
| with: | |
| tag: ${{ needs.docker_tag.outputs.tag }} | |
| status: ${{ needs.test.result }} | |
| secrets: inherit |