Document docker + hatch best practices

[thoughtfuldata]

Would be great if the hatch team could provide a how-to hatch with docker documentation

I am coming from poetry and wanted to try hatch.
Unfortunately, the documentation is light for a docker environment. I did notice there is a repo for hatch-containers but unsure if that is the best route.

Thanks!

[aiguofer]

This!! we just ran into a huge issue in production. We have a fastapi app that we run using uvicorn and wrap with DataDog. Our CMD is essentially:

hatch run ddtrace-run uvicorn app.main:app --host 0.0.0.0 --port 8000

This causes hatch to handle dependencies at runtime, which is NOT what you want. Something caused our dependency handling to exceed the Kubernetes liveliness timeout so it would just keep restarting the pod.

How should one set up a "prod" hatch env? it seems like we should use dev-mode = false... but then how do we make sure the deps are installed while building the Docker image? I can't seem to find a command to do this.

[doolio]

I think this discussion is mislabelled and should be categorised as Q&A.

[polarathene]

how do we make sure the deps are installed while building the Docker image?

If this is unrelated to the mentioned plugin (that uses a container for running the venv), but just a regular Dockerfile build of your own, this may be of interest to you.

That only focuses on a minimal image with hatch + uv, but from there you should be able to RUN hatch create env which should create the environment based on your pyproject.toml / hatch.toml deps. This way deps should be pulled in at the image build, so you should be good at runtime?