ci: cache efa installer once daily #52
Workflow file for this run
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: Package Generation | |
| on: [push, pull_request] | |
| jobs: | |
| srpm: | |
| name: "Generate SRPM" | |
| runs-on: ubuntu-latest | |
| container: | |
| image: fedora:rawhide | |
| options: --privileged | |
| steps: | |
| - run: dnf install -y packit git nodejs | |
| - uses: actions/checkout@v4 | |
| - run: git config --global --add safe.directory /__w/aws-ofi-nccl/aws-ofi-nccl | |
| - run: packit srpm | |
| - uses: actions/upload-artifact@v4 | |
| with: | |
| name: "libnccl-net-ofi.src.rpm" | |
| path: "*.src.rpm" | |
| if-no-files-found: 'error' | |
| compression-level: '0' | |
| mock-rpm: | |
| runs-on: ubuntu-latest | |
| container: | |
| image: fedora:39 | |
| options: --privileged --cap-add SYS_ADMIN | |
| needs: srpm | |
| strategy: | |
| matrix: | |
| include: | |
| - name: al2023-x86-cuda-aws | |
| args: "--define 'with_cuda 1' --define 'with_platform_aws 1'" | |
| accel_repo: --addrepo "https://developer.download.nvidia.com/compute/cuda/repos/amzn2023/x86_64/" | |
| efa_repo: "./aws-efa-installer/RPMS/ALINUX2023/x86_64/" | |
| mock: "/etc/mock/amazonlinux-2023-x86_64.cfg" | |
| - name: al2023-x86-neuron-aws | |
| args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| accel_repo: "" | |
| efa_repo: "./aws-efa-installer/RPMS/ALINUX2023/x86_64/" | |
| mock: "/etc/mock/amazonlinux-2023-x86_64.cfg" | |
| - name: rocky9-x86-neuron-aws | |
| args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| accel_repo: "" | |
| efa_repo: "./aws-efa-installer/RPMS/ROCKYLINUX9/x86_64/" | |
| mock: "/etc/mock/rocky+epel-9-x86_64.cfg" | |
| - name: rocky9-x86-cuda-aws | |
| args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| accel_repo: --addrepo "https://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/" | |
| efa_repo: "./aws-efa-installer/RPMS/ROCKYLINUX8/x86_64/" | |
| mock: "/etc/mock/rocky+epel-9-x86_64.cfg" | |
| - name: rocky8-x86-neuron-aws | |
| args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| accel_repo: "" | |
| efa_repo: "./aws-efa-installer/RPMS/ROCKYLINUX8/x86_64/" | |
| mock: "/etc/mock/rocky-8-x86_64.cfg" | |
| - name: rocky8-x86-cuda-aws | |
| args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| accel_repo: --addrepo "https://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/" | |
| efa_repo: "./aws-efa-installer/RPMS/ROCKYLINUX8/x86_64/" | |
| mock: "/etc/mock/rocky-8-x86_64.cfg" | |
| - name: leap155-x86-neuron-aws | |
| args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| accel_repo: "" | |
| efa_repo: "./aws-efa-installer/RPMS/SUSE/x86_64/" | |
| mock: "/etc/mock/opensuse-leap-15.5-x86_64.cfg" | |
| - name: leap155-x86-cuda-aws | |
| args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| accel_repo: --addrepo "https://developer.download.nvidia.com/compute/cuda/repos/opensuse15/x86_64/" | |
| efa_repo: "./aws-efa-installer/RPMS/SUSE/x86_64/" | |
| mock: "/etc/mock/opensuse-leap-15.5-x86_64.cfg" | |
| ## need to figure out binfmt_misc registration under gh actions for these to work. | |
| # - name: rocky9-aarch64-neuron-aws | |
| # args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| # accel_repo: "" | |
| # efa_repo: "./aws-efa-installer/RPMS/ROCKYLINUX9/aarch64/" | |
| # mock: "/etc/mock/rocky+epel-9-aarch64.cfg" | |
| # - name: rocky9-aarch64-cuda-aws | |
| # args: "--define 'with_neuron 1' --define 'with_platform_aws 1'" | |
| # accel_repo: --addrepo "https://developer.download.nvidia.com/compute/cuda/repos/rhel9/aarch64/" | |
| # efa_repo: "./aws-efa-installer/RPMS/ROCKYLINUX9/aarch64/" | |
| # mock: "/etc/mock/rocky+epel-9-aarch64.cfg" | |
| name: Build RPM for ${{ matrix.name }} | |
| steps: | |
| - run: dnf install -y mock nodejs createrepo_c | |
| - uses: actions/download-artifact@master | |
| with: | |
| name: "libnccl-net-ofi.src.rpm" | |
| path: . | |
| - uses: actions/cache@v3 | |
| with: | |
| key: aws-efa-installer-latest | |
| path: aws-efa-installer | |
| - run: bash -c "cd ${{ matrix.efa_repo }} && createrepo_c ." | |
| - run: mock --isolation=simple --verbose --addrepo file://$(realpath ${{ matrix.efa_repo }}) ${{ matrix.accel_repo }} -r ${{ matrix.mock }} ${{ matrix.args }} *.src.rpm | |
| - run: rpm -qpl /var/lib/mock/**/result/*.rpm | |
| - uses: actions/upload-artifact@v4 | |
| with: | |
| name: "RPMs - ${{ matrix.name }}" | |
| path: "/var/lib/mock/**/result/*.rpm" | |
| if-no-files-found: 'error' | |
| compression-level: '0' |