Call actions by name
Procedures are prepared as named sets you can call. The mechanism behind them is Zuul’s pipelines, jobs, and roles.
How pipelines, jobs, and roles fit together
- A pipeline is a flow of jobs: when each one runs, and in what order, is set
- A job is a set you call by name: parts (roles), the common processing that runs before and after, and the settings, combined. Everything in this guide — checks, measurement, delivery — is a job
- A role is a part inside a job: the steps themselves, bundled
How you use it
- Your project just calls a job by name (the same shape as calling an action with
uses:in CI) - You do not write the procedure in your own project. The procedure and its settings live on the called side
- When the contents change, every project that calls it gets the change. There is nothing to follow up per project
Against GitHub Actions
What you used to write in GitHub Actions, and what becomes of it on this base.
- checkout (fetching the code): the base fetches the change and prepares it in the workspace. The project writes nothing
- setup-node / setup-python (installing a language environment): the tools are prepared in the execution environment’s image, or a role prepares what is needed
- cache (caching dependencies): the place that makes things fast is on the base’s side (a shared store for dependencies, and the like). The project writes nothing
- upload-artifact (keeping outputs): results and evidence stay as the job’s output
- docker/build-push (building and publishing an image): a job that builds images and a job that publishes them carry this
- Deploy actions to a cloud: distribution is carried by a dedicated job (such as distributing an Android app)
- action-gh-release (creating a GitHub release): there is no dedicated counterpart