Ask HN: How do you provide your Python dependencies in production in 2016?

I have some python code that I want to use in production. The code depends on a few scientific packages and I want to add more (especially I'm going to use numba). Now I use virtualenv + pip + ansible role to install required system dependencies, but I've started to look to anaconda.

I'm in doubt about anaconda because it seems to be oriented more for development than for production usage. Does anybody use conda in production and is it good for it? What alternatives to it and virtualenv+pip are worth considering except docker?