I am facing a common challenge with Docker containers in a JupyterHub environment. Once a single user server container is spawned from an image, it’s essentially a snapshot of that image at that point in time. If I make changes to the image afterwards, those changes won’t automatically propagate to existing containers.
Any Ideas that i can keep Containers of single user servers uptodate?
In general containers are designed to provide a fixed reproducible environment. You could play around with persistent volumes and paths, e.g. to create a Conda/Python environment on persistent storage and modifying the image to use those paths instead of what’s built-in to the image, but there’s nothing that would let you persist the entire state of the container.
JupyterHub is too big, TLJH is overkill, best for a couple of friends to work on a safe jupyter environment would be, I guess, jupyter-service running at least one of docker-stack.
home at jovyan, there is .condarc to keep pkg cache,where I am now. If I can persist at least the package caches there,maybe I can spin-up containers way faster.