pyinfra automates service deployment. It does this by diff-ing the state of the server with the state defined in the deploy script. Deploys are asyncronous and highly performant. The inventory & deploy are managed with pure Python, allowing for near-infinite extendability.
When you run pyinfra you'll see something like:
pyinfra was designed from day one to enable ops to deploy things in a consistent, debuggable and maintainable manner. Notable design decisions:
- outputs shell commands and files to upload
- two-step deploy that enables dry-runs
- fail fast where possible (eg touching a directory)
-vmeans print out remote stdout & stderr in realtime- always print raw stderr on operation failure for instant debugging
- uses pure, 100% Python for the inventory and deploy scripts
- with operations/hooks to safely use Python mid-deploy
- properly agentless - even Python isn't required on the remote side (just a shell!)
pyinfra is still under heavy development, and while the CLI/API should be considered fairly
stable there's no guarantee of no breaking changes until v1. There are a number of critical
specifications to be properly fleshed out before the v1 release:
- spec/docs for roles/sub-deploys
- spec/docs for extension modules/facts
- spec/docs for extension deploys
To develop pyinfra itself:
# Create a virtualenv
venv create pyinfra
# Install pyinfra in dev mode, with dev requirements
pip install -e .[dev]Use nosetests or the bundled helper script to run tests. This helper script also counts
coverage:
# Test everything (API, modules & facts)
scripts/test.sh
# Set individual bits
scripts/test.sh [api|modules|facts]