CI: Unified multi-stage Docker build (x86, Ubuntu24)

Consolidates NIXL build procedures by using contrib/Dockerfile
multi-stage build for both CI and customer workflows.

Signed-off-by: Alexey Rivkin <[email protected]>

Author: Alexey-Rivkin

.ci/jenkins/lib/build-matrix.yaml

@@ -22,6 +22,10 @@
 ---
 job: nixl-ci-build
 
+registry_host: harbor.mellanox.com
+registry_auth: nixl_harbor_credentials
+registry_path: /nixl
+
 # Fail job if one of the steps fails or continue
 failFast: false
 
@@ -34,17 +38,30 @@ kubernetes:
   requests: "{memory: 8Gi, cpu: 8000m}"
 
 runs_on_dockers:
-  - { name: "ubuntu24.04-cuda-dl-base", url: "nvcr.io/nvidia/cuda-dl-base:25.06-cuda12.9-devel-ubuntu24.04" }
-  - { name: "ubuntu22.04-cuda-dl-base", url: "nvcr.io/nvidia/cuda-dl-base:24.10-cuda12.6-devel-ubuntu22.04" }
-  - { name: "podman-v5.0.2", url: "quay.io/podman/stable:v5.0.2", category: 'tool', privileged: true }
+  - {
+      file: "contrib/Dockerfile",
+      name: "ubuntu24.04-nixl-deps",
+      uri: "$arch/$name",
+      tag: "02112025",
+      build_args: "--target nixl-base"
+    }
+  # - {
+  #     file: "contrib/Dockerfile",
+  #     name: "ubuntu22.04-nixl-deps",
+  #     uri: "$arch/$name",
+  #     tag: "02112025",
+  #     build_args: "--target nixl-base --build-arg BASE_IMAGE_TAG=24.10-cuda12.6-devel-ubuntu22.04"
+  #   }
+  # - { name: "podman-v5.0.2", url: "quay.io/podman/stable:v5.0.2", category: 'tool', privileged: true }
 
 matrix:
   axes:
     arch:
       - x86_64
-      - aarch64
+      # - aarch64
 
 env:
+  NIXL_BASE_IMAGE_ENV: "true"
   NIXL_INSTALL_DIR: /opt/nixl
   TEST_TIMEOUT: 30
   UCX_TLS: "^shm"

.gitlab/build.sh

@@ -53,6 +53,12 @@ fi
 ARCH=$(uname -m)
 [ "$ARCH" = "arm64" ] && ARCH="aarch64"
 
+# Skip dependency installation if running in pre-built base image
+if [ -n "${NIXL_BASE_IMAGE_ENV}" ]; then
+    UCX_INSTALL_DIR=/usr
+    LIBFABRIC_INSTALL_DIR=/usr/local
+else
+
 # Some docker images are with broken installations:
 $SUDO rm -rf /usr/lib/cmake/grpc /usr/lib/cmake/protobuf
 
@@ -215,6 +221,8 @@ rm "libfabric-${LIBFABRIC_VERSION#v}.tar.bz2"
   cp gtest-parallel/* ${INSTALL_DIR}/bin/
 )
 
+fi # end NIXL_BASE_IMAGE_ENV check
+
 export LD_LIBRARY_PATH="${INSTALL_DIR}/lib:${INSTALL_DIR}/lib/$ARCH-linux-gnu:${INSTALL_DIR}/lib64:$LD_LIBRARY_PATH:${LIBFABRIC_INSTALL_DIR}/lib"
 export CPATH="${INSTALL_DIR}/include:${LIBFABRIC_INSTALL_DIR}/include:$CPATH"
 export PATH="${INSTALL_DIR}/bin:$PATH"

contrib/Dockerfile

@@ -17,7 +17,7 @@ ARG BASE_IMAGE="nvcr.io/nvidia/cuda-dl-base"
 ARG BASE_IMAGE_TAG="25.06-cuda12.9-devel-ubuntu24.04"
 ARG OS
 
-FROM ${BASE_IMAGE}:${BASE_IMAGE_TAG}
+FROM ${BASE_IMAGE}:${BASE_IMAGE_TAG} AS nixl-base
 
 # Set default OS if not provided
 ARG OS=${OS:-ubuntu24}
@@ -62,7 +62,28 @@ RUN apt-get update -y && \
     clang \
     hwloc \
     libhwloc-dev \
-    libcurl4-openssl-dev libssl-dev uuid-dev zlib1g-dev # aws-sdk-cpp dependencies
+    libcurl4-openssl-dev libssl-dev uuid-dev zlib1g-dev \
+    python3-dev \
+    python3-pip \
+    curl \
+    wget \
+    numactl \
+    git \
+    libiberty-dev \
+    libgoogle-glog-dev \
+    libgmock-dev \
+    libjsoncpp-dev \
+    libpython3-dev \
+    libboost-all-dev \
+    meson \
+    pkg-config \
+    pybind11-dev \
+    net-tools \
+    iproute2 \
+    pciutils \
+    libpci-dev \
+    libibmad-dev \
+    doxygen
 
 # Add DOCA repository and install packages
 RUN ARCH_SUFFIX=$(if [ "${ARCH}" = "aarch64" ]; then echo "arm64"; else echo "amd64"; fi) && \
@@ -167,6 +188,9 @@ RUN wget --tries=3 --waitretry=5 --timeout=30 --read-timeout=60 \
     make install && \
     ldconfig
 
+# Stage 2: Build NIXL (default stage)
+FROM nixl-base
+
 # By default, uv downloads python packages to $HOME/.cache/uv and hard links them
 # from the virtual environment. This means that the files reside in /root/.cache/uv,
 # which is not what we want since some systems mount user home dir into /root,
@@ -190,14 +214,20 @@ RUN uv pip install --upgrade meson meson-python pybind11 patchelf pyYAML click t
 RUN export UV_INDEX="https://download.pytorch.org/whl/cu$(echo $CUDA_VERSION | cut -d. -f1,2 | tr -d .)" && \
     uv pip install torch torchvision torchaudio
 
+# Re-declare ARGs needed in this stage
+ARG NIXL_PREFIX="/usr/local/nixl"
+ARG NIXL_PLUGIN_DIR="$NIXL_PREFIX/lib/x86_64-linux-gnu/plugins"
+ARG ARCH="x86_64"
+ARG DEFAULT_PYTHON_VERSION="3.12"
+ARG UCX_PREFIX="/usr"
+ARG UCX_PLUGIN_DIR="$UCX_PREFIX/lib/ucx"
+ARG LIBFABRIC_INSTALL_PATH="/usr/local"
+
 WORKDIR /workspace/nixl
 COPY . /workspace/nixl
 
 ENV LD_LIBRARY_PATH=/usr/local/lib:$LIBFABRIC_INSTALL_PATH/lib:$LD_LIBRARY_PATH
 
-# Install pybind11 via apt
-RUN apt-get update && apt-get install -y --no-install-recommends pybind11-dev
-
 ENV NIXL_PREFIX=$NIXL_PREFIX
 RUN rm -rf build && \
     mkdir build && \