Pipeline CI/CD no Azure DevOps (test, artifact, coverage gates)

Summary: I structure the pipeline in stages: Build → Test → Quality Gates → Pack → Deploy. Each stage must pass before the next runs.

1. Source & Build

Pull from main / feature branch.
Install dependencies (npm ci).
Build the application (Next.js / React) — outputs to dist/ or build/.

2. Test Stage

Unit tests: jest --coverage → coverage report (JUnit/HTML).
Integration tests: playwright test / cypress run.
Linting & formatting: eslint, prettier — fail the pipeline on violations.

Ver detalhes em testando componentes React.

3. Quality Gates (Azure DevOps Policy)

Code coverage gate: minimum 80% (configurable). Below it, pipeline fails.
Vulnerability scan: npm audit / snyk — blocks on critical/high vulns.
SonarQube: deeper static analysis. Gates: “No new bugs”, “Technical debt < 5%”.

4. Artifact Publishing

If all gates pass, the build artifact (e.g., .zip of dist/) is published to Azure Artifacts or a Container Registry.

5. Deploy Stage (environments)

Dev: automatic deployment after merge to develop.
Staging: manual approval trigger (or canary).
Production: manual approval + scheduled window.

6. Monitoring

Integrate Application Insights and New Relic to track performance after deployment. Automatic rollback triggers if error rates spike.

Real-world experience (Seguralta)

We used BitBucket Pipelines (similar to Azure DevOps). The coverage gate caught a 5% drop and prevented a release. The team fixed tests and improved coverage to 85% before shipping.

Relacionadas: versionamento da lib · índice.