CI/CD Pipeline Design | Blue/Green, Canary & GitOps

The Release Pipeline

A pipeline transforms code into a running service. It must be Idempotent and Immutable.

Code
(Git Push)

→

Build
(Docker Build)

→

Test
(Unit/Integration)

→

Release
(Push to Registry)

→

Deploy
(K8s Apply)

The Testing Pyramid

Don't just run E2E tests. They are slow and flaky. Balance your suite.

E2E (UI)
10% (Slow)

Integration (API)
20% (Medium)

Unit (Code)
70% (Fast)

Deployment Strategies

How do you update a running application without downtime?

Strategy	Description	Pros	Cons
Rolling Update	Replace N instances at a time (e.g., k8s default).	Zero downtime. No extra infrastructure cost.	Slow rollback. Hard to verify mid-flight.
Blue/Green	Spin up a full parallel environment (Green). Switch load balancer.	Instant rollback (Switch back to Blue). Safe.	Double infrastructure cost (2x resources).
Canary	Route 1% of traffic to new version. Gradually increase.	Real user testing. Lowest risk of widespread outage.	Complex routing (requires Service Mesh/LB).

Immutable Infrastructure

🚫 Mutable (Bad)

SSH into servers to run `git pull` or `apt-get install`. Leads to "Configuration Drift" and "Snowflake Servers".

✅ Immutable (Good)

Bake a new Docker Image or AMI for every change. Never patch running servers. If a server is sick, shoot it and spawn a new one.

Example: GitHub Actions Workflow

name: Production Build
on:
  push:
    tags: ["v*"]

jobs:
  build-and-push:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - name: Build Docker Image
        run: docker build -t myapp:${{ github.ref_name }} .
      - name: Push to Registry
        run: docker push myregistry.com/myapp:${{ github.ref_name }}
      - name: Deploy to K8s
        run: |
          # GitOps Pattern: Update the manifest repo, don't kubectl apply directly
          git clone [email protected]:org/ops-repo.git
          cd ops-repo
          yq -i '.spec.template.spec.containers[0].image = "myapp:${{ github.ref_name }}"' deployment.yaml
          git commit -am "Deploy v${{ github.ref_name }}"
          git push

Summary

Automate Everything: Humans should not touch production servers.
Fail Fast: Run unit tests first. If they fail, stop the pipeline immediately.
GitOps: The state of your cluster should mirror the state of your Git repository.