Hi Everyone.... I have a query regarding the pods of my project running on kubernetes which is using nvidia GPU's. I build my own custom image with a Dockerfile looks like this:

# Dockerfile
FROM <http://nvcr.io/nvidia/cuda:11.0.3-base-ubuntu20.04|nvcr.io/nvidia/cuda:11.0.3-base-ubuntu20.04>

# Set the working directory in the container
WORKDIR /app
ENV LANG C.UTF-8
ENV LC_ALL C.UTF-8
ENV PYTHONPATH /app

# Update the package list and install necessary packages
RUN apt-get update && \
    apt-get install -y python3 python3-venv python3-pip && \
    apt-get clean

# Copy the requirements file into the container
ENV VENV /opt/venv

COPY <http://requirements.in|requirements.in> .

# Create a virtual environment and install the required packages
RUN python3 -m venv ${VENV} && \
    . ${VENV}/bin/activate && \
    pip install --no-cache-dir -r <http://requirements.in|requirements.in>
    
ENV PATH="${VENV}/bin:$PATH"


COPY . .

# Alternatively, you can use ENTRYPOINT to automatically activate the virtual environment
# ENTRYPOINT ["/bin/bash", "-c", "source venv/bin/activate && exec bash"]
ARG tag
ENV FLYTE_INTERNAL_IMAGE $tag

It successfully runs with pyflyte run --remote --image="<build image>" workflows/main.py workflow_name And pods are using GPU's too. But Pods are running as a root instead of random pid. And It's safety concern for us. we can not allow to run any process as a root. Is there anything I did wrong in building image. Or any settings I have to do with flyte for this. Please help! Thank you

You have to set user in image refer to docker docs

@loud-keyboard-72751, for an example of how to run flytekit as non-root user, take a look at the official flytekit image: https://github.com/flyteorg/flytekit/blob/master/Dockerfile. Notice how we run as the non-root user

flytekit

in that case.

Thanks for this...It quite helps a lot.. i will try it and let you know...Thanks @freezing-airport-6809 too