fix: install rdma libs in runtime image. (#2163)

Signed-off-by: Kapil Arya <kapila@nvidia.com>

container/Dockerfile.sglang

@@ -58,6 +58,10 @@ RUN apt-get update -y && \
     meson \
     ninja-build \
     pybind11-dev \
+    # These headers are missing with the hpcx installer, required
+    # by UCX to find RDMA devices
+    libibverbs-dev rdma-core ibverbs-utils libibumad-dev \
+    libnuma-dev librdmacm-dev ibverbs-providers \
     # Rust build dependencies
 	clang \
     libclang-dev \
@@ -423,7 +427,15 @@ $LD_LIBRARY_PATH
 # libnuma-dev is a required dependency for sglang integration with NIXL
 COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/
 RUN apt-get update && \
-    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends build-essential python3-dev libnuma-dev && \
+    DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
+        build-essential python3-dev libnuma-dev \
+        # Curl for polling various endpoints.
+        curl \
+        # For debugging
+        vim \
+        # Libraries required by UCX to find RDMA devices
+        libibverbs1 rdma-core ibverbs-utils libibumad3 \
+        libnuma1 librdmacm1 ibverbs-providers && \
     rm -rf /var/lib/apt/lists/* && \
     uv venv $VIRTUAL_ENV --python 3.12 && \
     echo "source $VIRTUAL_ENV/bin/activate" >> ~/.bashrc

container/Dockerfile.tensorrt_llm

@@ -65,8 +65,14 @@ RUN apt update -y && \
     tmux \
     vim \
     ## NIXL dependencies
+    cmake \
     meson \
-    ninja-build
+    ninja-build \
+    pybind11-dev \
+    # These headers are missing with the hpcx installer, required
+    # by UCX to find RDMA devices
+    libibverbs-dev rdma-core ibverbs-utils libibumad-dev \
+    libnuma-dev librdmacm-dev ibverbs-providers
 
 ### UCX EFA Setup ###
 RUN rm -rf /opt/hpcx/ucx && \
@@ -374,6 +380,13 @@ RUN apt-get update && \
     DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
         build-essential \
         python3-dev \
+        # Curl for polling various endpoints.
+        curl \
+        # For debugging
+        vim \
+        # Libraries required by UCX to find RDMA devices
+        libibverbs1 rdma-core ibverbs-utils libibumad3 \
+        libnuma1 librdmacm1 ibverbs-providers \
         openssh-client \
         openssh-server && \
     rm -rf /var/lib/apt/lists/*

container/Dockerfile.vllm

@@ -56,6 +56,10 @@ RUN apt-get update -y && \
     meson \
     ninja-build \
     pybind11-dev \
+    # These headers are missing with the hpcx installer, required
+    # by UCX to find RDMA devices
+    libibverbs-dev rdma-core ibverbs-utils libibumad-dev \
+    libnuma-dev librdmacm-dev ibverbs-providers \
     # Rust build dependencies
 	clang \
     libclang-dev \
@@ -72,10 +76,6 @@ RUN apt-get update -y && \
     automake \
     libtool \
     net-tools \
-    # These headers are missing with the hpcx installer, required
-    # by UCX to find RDMA devices
-    libibverbs-dev rdma-core ibverbs-utils libibumad-dev \
-    libnuma-dev librdmacm-dev ibverbs-providers \
     # For Prometheus
     curl tar ca-certificates && \
     rm -rf /var/lib/apt/lists/*
@@ -437,6 +437,13 @@ RUN apt-get update && \
     DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
         build-essential \
         python3-dev \
+        # Curl for polling various endpoints.
+        curl \
+        # For debugging
+        vim \
+        # Libraries required by UCX to find RDMA devices
+        libibverbs1 rdma-core ibverbs-utils libibumad3 \
+        libnuma1 librdmacm1 ibverbs-providers \
         # JIT Kernel Compilation, flashinfer
         ninja-build \
         g++ \