This document describes the requirements for committing to this repository.
In order to contribute to this project, you must sign each of your commits to attest that you have the right to contribute that code.
This is done with the -s
/--signoff
flag on git commit
.
More information about DCO
can be found here
All code that is contributed to kube-rs must go through the Pull Request (PR) process. To contribute a PR, fork this project, create a new branch, make changes on that branch, and then use GitHub to open a pull request with your changes.
Every PR must be reviewed by at least one Maintainer of the project. Once a PR has been marked "Approved" by a Maintainer (and no other Maintainer has an open "Rejected" vote), the PR may be merged. While it is fine for non-maintainers to contribute their own code reviews, those reviews do not satisfy the above requirement.
This project has adopted the CNCF Code of Conduct.
- Channel: Code is built and tested using the stable channel of Rust, but documented and formatted with nightly *
- Formatting: To format the codebase, run
make fmt
- Documentation To check documentation, run
make doc
- Testing: To run tests, run
make test
and see below.
For a list of tooling that we glue together everything see TOOLS.md.
We have 3 classes of tests.
- Unit tests & Documentation Tests
- Integration tests (requires Kubernetes)
- End to End tests (requires Kubernetes)
The last two will try to access the Kubernetes cluster that is your current-context
; i.e. via your local KUBECONFIG
evar or ~/.kube/config
file.
The easiest way set up a minimal Kubernetes cluster for these is with k3d
(make k3d
).
Most unit/doc tests are run from cargo test --lib --doc --all
, but because of feature-sets, and examples, you will need a couple of extra invocations to replicate our CI.
For the complete variations, run the make test
target in the Makefile
.
All public interfaces must be documented, and most should have minor documentation examples to show usage.
Slower set of tests within the crates marked with an #[ignore]
attribute.
Most integration tests are run with cargo test --all --lib -- --ignored
, but because of feature-sets, you will need a few invocations of these to replicate our CI. See make test-integration
We have a small set of e2e tests that tests difference between in-cluster and local configuration.
These tests are the heaviest tests we have because they require a full docker build
, image import (or push/pull flow), yaml construction, and kubectl
usage to verify that the outcome was sufficient.
To run E2E tests, use (or follow) make e2e
as appropriate.
All public interfaces should have doc tests with examples for docs.rs.
When adding new non-trivial pieces of logic that results in a drop in coverage you should add a test.
Cross-reference with the coverage build and go to your branch. Coverage can also be run locally with cargo tarpaulin
at project root. This will use our tarpaulin.toml config, and will run both unit and integration tests.
- Unit tests MUST NOT try to contact a Kubernetes cluster
- Doc tests MUST be marked as
no_run
when they need to contact a Kubernetes cluster - Integration tests MUST NOT be used when a unit test is sufficient
- Integration tests MUST NOT assume existence of non-standard objects in the cluster
- Integration tests MUST NOT cross-depend on other unit tests completing (and installing what you need)
- E2E tests MUST NOT be used where an integration test is sufficient
In general: use the least powerful method of testing available to you:
- use unit tests in
kube-core
- use unit tests in
kube-client
(and in rare cases integration tests) - use unit tests in
kube-runtime
(and occassionally integration tests) - use e2e tests when testing differences between in-cluster and local configuration
The high-level architecture document is written for contributors.
You can ask general questions / share ideas / query the community at the kube-rs discussions forum. You can reach the maintainers of this project at #kube channel on the Tokio discord.