Promoting Code with Pipeline Stages

Why it exists

To solve the problem of shipping code directly from a developer's machine to production. Without a structured promotion process, you risk deploying broken, untested, or insecure code, leading to outages. Staged promotion creates a predictable, repeatable, and safe path for software to reach users.

The mental model

A pipeline is an assembly line for your code. Each stage is a workstation on that line. The 'Build' stage assembles the parts. The 'Test' stage runs quality checks. The 'Staging' stage does a final dress rehearsal. The 'Production' stage is the final delivery to the customer. You don't ship a car from the first workstation; you promote it down the line only after it passes inspection at each step.

How it works

In a CI/CD tool like Azure Pipelines, you define a series of stages. Each stage is a logical boundary containing one or more jobs, like building code or running tests. By default, stages run sequentially; Stage 'B' only starts if Stage 'A' succeeds. This creates a promotion model where an artifact must pass the checks of one stage to be eligible for the next. You can also add conditions, dependencies, and gates. For example, promotion to the 'Production' stage might require manual approval from a manager, while promotion from 'Build' to 'Test' is fully automatic upon success.

When to use it

Use a multi-stage pipeline for any application that requires more than a simple build-and-test. It's essential for any service with distinct environments (e.g., staging, production). This structure is the foundation of modern CI/CD, providing visibility and control over how code moves from commit to customer.

When not to use it

For very simple projects, like a small command-line tool or a script that doesn't get 'deployed' to an environment, a single-stage pipeline with just build and test jobs is sufficient. Adding multiple stages can be overkill if there are no distinct environments or deployment processes to model.

One canonical example

A common four-stage pipeline for a web service: First, the 'Build' stage compiles the code and runs static analysis. Second, the 'Test' stage runs unit and integration tests against the built artifact. Third, the 'Staging' stage deploys the artifact to a staging environment that mirrors production and runs end-to-end tests. Fourth, the 'Production' stage deploys the same verified artifact to users, often after a manual approval gate.