Design and Implement CI/CD Deployment Pipelines

Deployment Pipeline Design provides architecture patterns for multi-stage CI/CD pipelines with approval gates and deployment strategies, aiming to balance speed with safety. It defines a standard nine-stage pipeline flow - Source, Build, Test, Staging Deploy, Integration Tests, Approval Gate, Production Deploy, Verification, and Rollback - and three approval-gate patterns: Manual Approval (GitHub Actions environment protection rules), Time-Based Approval (GitLab CI's when: delayed with a start_in wait), and Multi-Approver (Azure Pipelines' ManualValidation task notifying named reviewers). Four deployment strategies are documented with characteristics: Rolling Deployment (gradual, zero-downtime, easy rollback, best for most apps) via Kubernetes' RollingUpdate strategy:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: my-app
spec:
  replicas: 10
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxSurge: 2
      maxUnavailable: 1

Blue-Green (instant switchover, easy rollback, temporarily doubles infrastructure cost, good for high-risk deploys), Canary (gradual traffic-weight shifts with pauses via Argo Rollouts, risk mitigation with real user testing, requires a service mesh), and Feature Flags (deploy without releasing, A/B testing, instant rollback, illustrated with the Flagsmith Python SDK).

Use this skill when designing CI/CD architecture, implementing deployment gates, configuring multi-environment pipelines, establishing deployment best practices, or implementing progressive delivery. It is not for tasks unrelated to deployment pipeline design or for work in a different domain or tool outside this scope.

Given a pipeline-design task, the skill outputs a full multi-stage GitHub Actions pipeline example - build, test with a Trivy security scan, deploy-staging, integration-test, canary deploy-production via Argo Rollouts promote, and a verify job with a health check and Slack webhook notification - plus rollback implementations: an automated rollback that health-checks after deploy and runs kubectl rollout undo on failure, and manual rollback commands (kubectl rollout history, kubectl rollout undo --to-revision=N). It also tracks six key pipeline metrics - Deployment Frequency, Lead Time, Change Failure Rate, Mean Time to Recovery (MTTR), Pipeline Success Rate, and Average Pipeline Duration - and a Prometheus-based post-deployment error-rate check that fails the pipeline if the error rate exceeds a threshold.

References github-actions-templates, gitlab-ci-patterns, and secrets-management as related skills for deeper implementation in each ecosystem, plus references/pipeline-orchestration.md for complex pipeline patterns and assets/approval-gate-template.yml for reusable approval-workflow templates. Ten best practices are stated: fail fast by running quick tests first, run independent jobs in parallel, cache dependencies between runs, manage build artifacts, keep environments at parity, use a secrets store such as Vault, schedule deployment windows appropriately, integrate monitoring, automate rollback on failure, and document pipeline stages.

Platform and DevOps engineers designing or hardening a CI/CD pipeline who need concrete, working examples of approval gates, deployment strategies, and rollback automation across GitHub Actions, GitLab CI, Azure Pipelines, and Kubernetes/Argo Rollouts, rather than assembling a pipeline from scratch.