Basically, trying to figure out if there is better way or what the best practice is for my use case. I'll try my best to explain this.

• I have a Django backend where I am using virtualenv to separate the app dependencies from the OS on my computer
• All of the secrets for the Django app (SECRET_KEY, db credentials, API keys) are in the virtualenv environment and read into the app with os.environ(<secret>)
• The code is eventually moved into a Docker image using python:3.7-slim where I do not use virtualenv because I don't need to isolate the app dependencies from the image OS
• I'm spinning up the Kubernetes cluster with Skaffold locally which pulls the image so I can do local development

So all that being said, the issue I run into is the following and why I still need to use virtualenv. Maybe it isn't an issue, but just strikes me as "irregular".

Sometimes I need to execute commands for Django:

python manage.py collectstatic
python manage.py migrate
python manage.py makemigrations
etc...

Some of these are fine to run with:

kubectl exec -it deploy/server-deployment -- python manage.py <command>

However, something like is not ok to run:

kubectl exec -it deploy/server-deployment -- python manage.py makemigrations

Why? Because the makemigrations will be generated in the Pod which will be lost when the Pod is destroyed. Thus, I need to run makemigrations locally so that I can push the migrations to Git, and doing it locally will trigger the images and containers to be rebuilt with the migrations in them.

That is why I still need to run virtualenv locally: so I can run Django commands locally.

If I do not run virtualenv locally two things will happen:

• App dependencies will be mixed with OS instead of isolated
• The program will fail because os.environ(<secret>) references will be empty in the app

So it seems like I have to run virtualenv locally and the thing that feels "wrong" to me is that I have to set environmental variables in two places:

1. virtualenv where I just put them in the ./bin/activate
2. k8s with kubectl create secret

I can automate this with a script, but still feels like double entry.

What is the correct way or best practice for doing this?

I hope that makes sense.