What is best practice for passing files and directories as command line arguments to a containerized program?

Question

I am working to containerize a program. The program requires a configuration file and directories for its operation, which are passed as command line arguments, e.g. something like this:

$ myprog -c /path/to/config.conf -i /directory/with/input_files -o /directory/with/output_files

I know that I will need mount docker volumes to the containerized version of this program so that it can access the config file and directories. But, I would like to maintain the command-line interface unchanged (as far as possible) to hide the complexity of using docker volumes, etc. from users. How do I do this?

Will I need to create a wrapper script that maps the values of the command-line arguments (-c, -i, -o) to mount the corresponding paths as docker volumes? If so, what is the best practice for packaging it (i.e. to distribute the script with the container) and making this work cross-platform (i.e. so that it runs on Windows, MacOS, Linux)?

Answer 1

I would suggest the following approach for the files within your docker-compose.yml:

volumes:
    - ./config:/path/to
    - ./input:/directory/with/input_files
    - ./output:/directory/with/output_files

Do note that I mount the entire folder in this example, despite you only mentioning one config file. This is on purpose - reasons for that are well explained within this video. Resources on how to define docker-compose.yml in full can be found here. Your final approach could look something like this:

1. Push Dockerfile to registry of your choice
2. Provide user the docker-compose.yml for example via Git
3. User creates folder with the necessary sub-directories (config, input and output in this example)
4. User supplies valid config file in config folder
5. User executes docker-compose up -d
6. Success :)