I am building containers that require the software stack necessary to support the building and running of JEDI applications in addition to some other stuff for building other capabilities. I’d like to use a multi-stage build of the container to copy the required bits and pieces from
jcsda/docker-gnu-openmpi-dev:latest, for example. Due to conflicts between conda-based software installs that I need for my project and the Spack environment used by JCSDA’s containers on Docker hub, this is proving to be challenging. It would be a lot easier to see what is going on if I knew how JCSDA is creating the containers it is publishing on Docker Hub. The
/root/entrypoint.sh script and the hard-coding of the Spack environment it sources, make it difficult to use JCSDA’s container as a base layer to mix in additional software on top of it.
Can anyone provide a pointer to the Dockerfile used to create those containers, or some other recipe so that I can better see what is going on? It would be very helpful to know exactly how those containers are being created.