Python Deployments Wishlist

Jun 29, 2018

Am currently faced with a baffling realisation of the reality that there is no one sane way to deploy python applications.

It’s utterly annoying that I cannot just tar a virtual env and use it.

Performing pip install on every deploy is plain ridiculous. Downloading from the internet on every code change with separate virtualenv is a nogo. No RPMs, nope, no way am writing another spec file.

Wishlist:

• Deployment tooling that allows me to go from zero to production with a new application with no setup.
• No-downtime between deployments (has to work with a single instance)
• Philosophy of implementations should make sense whether deploying to one or a 100 servers.
• Controlled activation of the new version of software i.e. traffic shifting.
• Stable re-producible environment setup everytime (perhaps with ansible).
• Cli that lets you bootstrap, build, deploy and control activation of software (nginx or via LB). Next:

Am starting to hack a prototype with fabric, python and Ansible, which is already proving to be challenging but looks doable.

More posts

• Next: Books in 2018
• Previous: Bug in my code and Ballet