Docker Container Existing after run with no error in log

Question

I have a python app consisting of image analysis models and 2 script files. In Main.py I have XMLRPC server to run forever listening to the clients.

if __name__ == "__main__":
    server = SimpleXMLRPCServer(("0.0.0.0", 8888))
    print("Listening on port 8888...")
    server.register_function(result, "result")
    server.serve_forever()

My Dcokerfile is:

# Start with NVIDIA's CUDA and cuDNN base image.
FROM nvidia/cuda:8.0-cudnn5-devel-ubuntu16.04

# Argument: the username & password.
ARG username
ARG user_password

# Update the system.
RUN echo "debconf debconf/frontend select Noninteractive" | debconf-set-selections
RUN apt-get update
RUN apt-get upgrade --assume-yes

...... bla bla bla

WORKDIR /home/${username}

# Copy the current directory contents into the container at /home/${username}
ADD . /home/${username}

...... bla bla bla

# Expose the ports and start the ssh daemon as entry point.
USER root
EXPOSE 22 6006 8888
ENTRYPOINT ["/usr/sbin/sshd", "-D"]

When I add CMD to run my Main.py The container does not work, It exiting immediately. What is the best practice which I be able to Run this container? I am using azure Data Science Virtual Machine for Linux Ubuntu.

I built my Dockerfile with:

 docker build . --tag img_processing:V1 --build-arg "username=blabla" --build-arg "user_password=blabla"

And I Run my Container with:

docker run -d -p 4000:8888 img_processing

Currently I use docker exec -it my-app-container bash and inside of my container I manage stuff and Run python Main.py & to run the script in background which I don't think is a good way. especially I have to find the way to scale up and process 3000 images at time. So each container needs to have same setup.

Any Idea?

Answer 1

First of all never expose PORT 22 and run SSH inside containers. It is not a recommended thing to do

Next you can specify the either ENTRYPOINT or CMD as

CMD ["python", "Main.py"]

For this you need to make sure that Main.py is in the current directory of the WORKDIR that you have specified

You would only use ENTRYPOINT when you want that argument passed to your container run command are appended to the ENTRYPOINT command. In your case CMD should do.

When your contain exits immediately you should remove the -d flag and use -it flag to debug the issue

docker run -it -p 4000:8888 img_processing

Answer 2

You can do ENTRYPOINT ["python", "Main.py"]. This will set python Main.py to be the default command of the container. You can read more abou ENTRYPOINT in the Dockerfile documentation.