Skip to content

fix(docker): set LD_LIBRARY_PATH for AWS EFA OFI plugin discovery #2585

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
terrykong merged 1 commit into from
May 29, 2026
Merged

fix(docker): set LD_LIBRARY_PATH for AWS EFA OFI plugin discovery #2585

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 11 additions & 0 deletions docker/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -211,6 +211,17 @@ LABEL com.nvidia.build.ref="${NVIDIA_BUILD_REF}"

ENV NEMO_RL_VENV_DIR=/opt/ray_venvs

# AWS EFA OFI plugin discovery (p4d / p5 / p5en multi-node NCCL via SRD).
# The host's aws-efa-installer places libnccl-net-ofi.so under
# /opt/amazon/ofi-nccl/lib and libfabric under /opt/amazon/efa/lib. Without
# these on the linker search path, NCCL silently falls back to its socket
# transport. Register both via ld.so.conf.d (so `ldconfig` populates the cache
# once host paths are mounted at runtime) and LD_LIBRARY_PATH (so NCCL's
# dlopen resolves the plugin even before `ldconfig` is re-run).
RUN printf '%s\n' /opt/amazon/ofi-nccl/lib /opt/amazon/efa/lib \
> /etc/ld.so.conf.d/aws-efa.conf
ENV LD_LIBRARY_PATH="/opt/amazon/ofi-nccl/lib:/opt/amazon/efa/lib:${LD_LIBRARY_PATH}"

# Copy in source from build context (defaults to cloned repo, can be overridden)
# Exclude pyproject.toml and uv.lock since those may be altered by build-custom-vllm.sh
COPY --from=nemo-rl --exclude=pyproject.toml --exclude=uv.lock . /opt/nemo-rl
Expand Down
11 changes: 11 additions & 0 deletions docker/Dockerfile.ngc_pytorch
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -139,6 +139,17 @@ LABEL com.nvidia.build.ref="${NVIDIA_BUILD_REF}"

ENV NEMO_RL_VENV_DIR=/opt/ray_venvs

# AWS EFA OFI plugin discovery (p4d / p5 / p5en multi-node NCCL via SRD).
# The host's aws-efa-installer places libnccl-net-ofi.so under
# /opt/amazon/ofi-nccl/lib and libfabric under /opt/amazon/efa/lib. Without
# these on the linker search path, NCCL silently falls back to its socket
# transport. Register both via ld.so.conf.d (so `ldconfig` populates the cache
# once host paths are mounted at runtime) and LD_LIBRARY_PATH (so NCCL's
# dlopen resolves the plugin even before `ldconfig` is re-run).
RUN printf '%s\n' /opt/amazon/ofi-nccl/lib /opt/amazon/efa/lib \
> /etc/ld.so.conf.d/aws-efa.conf
ENV LD_LIBRARY_PATH="/opt/amazon/ofi-nccl/lib:/opt/amazon/efa/lib:${LD_LIBRARY_PATH}"

# Copy in source from build context (defaults to cloned repo, can be overridden)
COPY --from=nemo-rl . /opt/nemo-rl
# Unshallow the repo to get the full history (in the case it was from the scratch layer).
Expand Down
Loading