|
| 1 | +# Base Dockerfile for NIXL CI/CD system |
| 2 | +# Contains shared dependencies: system packages, UCX, Rust, AWS SDK, etc. |
| 3 | +# Used by both NIXL and NIXLBench builds |
| 4 | + |
| 5 | +ARG BASE_IMAGE="nvcr.io/nvidia/cuda-dl-base" |
| 6 | +ARG BASE_IMAGE_TAG="25.06-cuda12.9-devel-ubuntu24.04" |
| 7 | +ARG OS="ubuntu24.04" |
| 8 | +ARG ARCH="x86_64" |
| 9 | +ARG DEFAULT_PYTHON_VERSION="3.12" |
| 10 | +ARG UCX_REF="v1.19.x" |
| 11 | +ARG NPROC="16" |
| 12 | + |
| 13 | +FROM ${BASE_IMAGE}:${BASE_IMAGE_TAG} |
| 14 | + |
| 15 | +ARG ARCH |
| 16 | +ARG OS |
| 17 | +ARG DEFAULT_PYTHON_VERSION |
| 18 | +ARG NPROC |
| 19 | +ARG UCX_REF |
| 20 | + |
| 21 | +# Install common system packages, build tools, and RDMA/InfiniBand packages |
| 22 | +RUN apt-get update -y && \ |
| 23 | + apt-get install -y ubuntu-keyring && \ |
| 24 | + DEBIAN_FRONTEND=noninteractive apt-get -y install \ |
| 25 | + autotools-dev \ |
| 26 | + automake \ |
| 27 | + build-essential \ |
| 28 | + cmake \ |
| 29 | + etcd-client \ |
| 30 | + etcd-server \ |
| 31 | + flex \ |
| 32 | + ibverbs-providers \ |
| 33 | + ibverbs-utils \ |
| 34 | + libaio-dev \ |
| 35 | + libibumad-dev \ |
| 36 | + libibverbs-dev \ |
| 37 | + libclang-dev \ |
| 38 | + libcpprest-dev \ |
| 39 | + libcurl4-openssl-dev \ |
| 40 | + libgflags-dev \ |
| 41 | + libgrpc-dev \ |
| 42 | + libgrpc++-dev \ |
| 43 | + libgtest-dev \ |
| 44 | + libnuma-dev \ |
| 45 | + libprotobuf-dev \ |
| 46 | + librdmacm-dev \ |
| 47 | + libssl-dev \ |
| 48 | + libtool \ |
| 49 | + liburing-dev \ |
| 50 | + libz-dev \ |
| 51 | + ninja-build \ |
| 52 | + protobuf-compiler-grpc \ |
| 53 | + pybind11-dev \ |
| 54 | + python${DEFAULT_PYTHON_VERSION}-dev \ |
| 55 | + rdma-core \ |
| 56 | + uuid-dev \ |
| 57 | + zlib1g-dev |
| 58 | + |
| 59 | +# Add Mellanox repository and install DOCA packages |
| 60 | +RUN case "${ARCH}" in \ |
| 61 | + aarch64) ARCH_SUFFIX="arm64-sbsa" ;; \ |
| 62 | + x86_64) ARCH_SUFFIX="${ARCH}" ;; \ |
| 63 | + esac && \ |
| 64 | + export PKG_CONFIG_PATH="/opt/mellanox/doca/lib/${ARCH_SUFFIX}-linux-gnu/pkgconfig:/opt/mellanox/dpdk/lib/${ARCH_SUFFIX}-linux-gnu/pkgconfig:$PKG_CONFIG_PATH" && \ |
| 65 | + curl -fsSL https://linux.mellanox.com/public/repo/doca/3.0.0/${OS}/${ARCH_SUFFIX}/GPG-KEY-Mellanox.pub | \ |
| 66 | + gpg --dearmor | tee /usr/share/keyrings/mellanox-archive-keyring.gpg && \ |
| 67 | + echo "deb [signed-by=/usr/share/keyrings/mellanox-archive-keyring.gpg] https://linux.mellanox.com/public/repo/doca/3.0.0/${OS}/${ARCH_SUFFIX} ./" | \ |
| 68 | + tee /etc/apt/sources.list.d/mellanox.list && \ |
| 69 | + apt-get update -y && \ |
| 70 | + apt-get install -y --no-install-recommends \ |
| 71 | + doca-all \ |
| 72 | + doca-sdk-common doca-sdk-dma doca-sdk-dpdk-bridge \ |
| 73 | + doca-sdk-eth doca-sdk-flow doca-sdk-rdma \ |
| 74 | + mlnx-dpdk mlnx-dpdk-dev |
| 75 | + |
| 76 | +# Install Ubuntu 24 specific DOCA packages |
| 77 | +RUN if [ "$OS" = "ubuntu24.04" ]; then \ |
| 78 | + apt-get install -y --no-install-recommends \ |
| 79 | + doca-sdk-gpunetio \ |
| 80 | + libdoca-sdk-gpunetio-dev; \ |
| 81 | + fi |
| 82 | + |
| 83 | +# Install AWS CLI |
| 84 | +RUN curl "https://awscli.amazonaws.com/awscli-exe-linux-${ARCH}.zip" -o "awscliv2.zip" && \ |
| 85 | + unzip awscliv2.zip && ./aws/install && rm -rf awscliv2.zip aws |
| 86 | + |
| 87 | +WORKDIR /workspace |
| 88 | + |
| 89 | +# Build and install etcd-cpp-apiv3 |
| 90 | +RUN git clone https://github.com/etcd-cpp-apiv3/etcd-cpp-apiv3.git && \ |
| 91 | + cd etcd-cpp-apiv3 && \ |
| 92 | + sed -i '/^find_dependency(cpprestsdk)$/d' etcd-cpp-apiv3-config.cmake.in && \ |
| 93 | + sed -i '/^find_dependency(cpprestsdk)$/d' cmake/etcd-cpp-apiv3-config.cmake.in && \ |
| 94 | + mkdir build && cd build && \ |
| 95 | + cmake .. -DCMAKE_INSTALL_PREFIX=/usr/local && \ |
| 96 | + make -j${NPROC} && \ |
| 97 | + make -j${NPROC} install |
| 98 | + |
| 99 | +# Build and install AWS SDK C++ |
| 100 | +RUN git clone --recurse-submodules https://github.com/aws/aws-sdk-cpp.git --branch 1.11.581 && \ |
| 101 | + mkdir aws_sdk_build && cd aws_sdk_build && \ |
| 102 | + cmake ../aws-sdk-cpp/ -DCMAKE_BUILD_TYPE=Release -DBUILD_ONLY="s3" -DENABLE_TESTING=OFF -DCMAKE_INSTALL_PREFIX=/usr/local && \ |
| 103 | + make -j${NPROC} && \ |
| 104 | + make -j${NPROC} install |
| 105 | + |
| 106 | +# Install uv for Python virtual environment management |
| 107 | +COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/ |
| 108 | + |
| 109 | +# Build and install UCX |
| 110 | +RUN rm -rf /usr/lib/ucx /opt/hpcx/ucx && \ |
| 111 | + cd /usr/local/src && \ |
| 112 | + git clone https://github.com/openucx/ucx.git && \ |
| 113 | + cd ucx && \ |
| 114 | + git checkout $UCX_REF && \ |
| 115 | + ./autogen.sh && ./configure \ |
| 116 | + --prefix=/usr \ |
| 117 | + --with-rdmacm \ |
| 118 | + --with-verbs \ |
| 119 | + --with-mlx5-dv \ |
| 120 | + --with-dm \ |
| 121 | + --without-java \ |
| 122 | + --enable-optimizations \ |
| 123 | + --disable-logging \ |
| 124 | + --disable-debug \ |
| 125 | + --disable-assertions \ |
| 126 | + --disable-params-check \ |
| 127 | + --enable-mt && \ |
| 128 | + make -j${NPROC} && \ |
| 129 | + make -j${NPROC} install-strip && \ |
| 130 | + ldconfig |
| 131 | + |
| 132 | +# Install Rust toolchain |
| 133 | +ENV RUSTUP_HOME=/usr/local/rustup \ |
| 134 | + CARGO_HOME=/usr/local/cargo \ |
| 135 | + PATH=/usr/local/cargo/bin:$PATH \ |
| 136 | + RUST_VERSION=1.86.0 |
| 137 | + |
| 138 | +RUN case "${ARCH}" in \ |
| 139 | + aarch64) RUSTARCH="aarch64-unknown-linux-gnu" ;; \ |
| 140 | + x86_64) RUSTARCH="x86_64-unknown-linux-gnu" ;; \ |
| 141 | + esac && \ |
| 142 | + wget --tries=3 --waitretry=5 \ |
| 143 | + "https://static.rust-lang.org/rustup/archive/1.28.1/${RUSTARCH}/rustup-init" \ |
| 144 | + "https://static.rust-lang.org/rustup/archive/1.28.1/${RUSTARCH}/rustup-init.sha256" && \ |
| 145 | + sha256sum -c rustup-init.sha256 && \ |
| 146 | + chmod +x rustup-init && \ |
| 147 | + ./rustup-init -y --no-modify-path --profile minimal --default-toolchain $RUST_VERSION --default-host ${RUSTARCH} && \ |
| 148 | + rm rustup-init* && \ |
| 149 | + chmod -R a+w $RUSTUP_HOME $CARGO_HOME |
| 150 | + |
| 151 | +# Final cleanup |
| 152 | +RUN rm -rf etcd-cpp-apiv3 aws-sdk-cpp aws_sdk_build /usr/local/src/ucx && \ |
| 153 | + apt-get clean && rm -rf /var/lib/apt/lists/* |
| 154 | + |
| 155 | +# Set up environment variables |
| 156 | +ENV LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH |
| 157 | +ENV PKG_CONFIG_PATH=/usr/local/lib/pkgconfig:$PKG_CONFIG_PATH |
0 commit comments