Delivery & Infrastructure

Guidelines for cloud architecture, CI/CD, and deployment engineering.

When to Use

• Designing cloud infrastructure (AWS/Azure/GCP)
• Setting up CI/CD pipelines
• Container orchestration (Kubernetes)
• Infrastructure as Code (Terraform)
• Deployment strategy selection

Core Principles

1. Automate Everything - No manual deployment steps
2. Infrastructure as Code - Version-controlled, reproducible
3. Build Once, Deploy Anywhere - Immutable artifacts
4. Fast Feedback - Fail fast, fix fast
5. Security by Design - Embedded throughout lifecycle
6. GitOps as Truth - Git is single source of truth
7. Zero-Downtime - All deployments without user impact

CI/CD Pipeline

Standard Stages

Lint → Test → Security Scan → Build → Deploy (staging) → Deploy (prod)

Pipeline Checklist

• Parallel execution where possible
• Caching for faster builds
• Environment-based deployment gates
• Automated rollback on failure
• Security scanning (SAST/DAST)
• Artifact versioning and signing

GitHub Actions Example

name: CI/CD
on:
  push:
    branches: [main]
  pull_request:
    branches: [main]

jobs:
  lint:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: npm run lint

  test:
    needs: lint
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: npm ci
      - run: npm test -- --coverage

  build:
    needs: test
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: docker/build-push-action@v5
        with:
          push: true
          tags: ${{ env.IMAGE }}:${{ github.sha }}

Container Best Practices

Dockerfile Checklist

• Multi-stage builds for minimal size
• Non-root user for runtime
• No secrets in image layers
• Minimal base image (alpine, distroless)
• Health checks defined
• Resource limits configured

Example Dockerfile

FROM node:20-alpine AS builder
WORKDIR /app
COPY package*.json ./
RUN npm ci --only=production
COPY . .
RUN npm run build

FROM node:20-alpine
RUN adduser -S app -u 1001
WORKDIR /app
COPY --from=builder --chown=app /app/dist ./dist
COPY --from=builder --chown=app /app/node_modules ./node_modules
USER app
EXPOSE 3000
HEALTHCHECK CMD node healthcheck.js
CMD ["node", "dist/main.js"]

Deployment Strategies

kubectl rollout undo

Kubernetes Essentials

Production Deployment Checklist

• Resource requests and limits defined
• Liveness and readiness probes
• Security context (non-root, read-only)
• Pod anti-affinity for HA
• HPA for auto-scaling
• PDB for availability during updates
• Secrets from external manager

Key Resources

# Deployment with probes
livenessProbe:
  httpGet:
    path: /health
    port: 3000
  initialDelaySeconds: 30
readinessProbe:
  httpGet:
    path: /ready
    port: 3000
  initialDelaySeconds: 10

Infrastructure as Code

Terraform Best Practices

• Use modules for reusable components
• Remote state with locking
• Input validation on variables
• Semantic versioning for modules
• Consistent naming conventions
• Document all outputs

Module Structure

modules/
  vpc/
    main.tf
    variables.tf
    outputs.tf

Cloud Architecture

Design Principles

1. Design for Failure - Self-healing mechanisms
2. Security by Design - Least privilege, encryption
3. Cost-Conscious - Right-size, savings plans
4. Managed Services - Reduce operational overhead

HA/DR Patterns

• Multi-AZ for availability
• Multi-region for disaster recovery
• Cross-region backup replication
• DNS-based failover

Cost Optimization

• Right-sizing based on usage
• Reserved instances/savings plans
• Lifecycle policies for storage
• Auto-scaling to match demand
• Regular unused resource cleanup

Security Checklist

• VPC with private subnets
• Security groups (least privilege)
• IAM roles (minimal permissions)
• Encryption at rest and in transit
• Secrets in dedicated manager
• Audit logging enabled
• DDoS protection configured

Observability

Metrics to Track

• RED: Rate, Errors, Duration
• USE: Utilization, Saturation, Errors

Key Dashboards

• Application performance (latency, errors)
• Infrastructure health (CPU, memory, disk)
• Business metrics (transactions, users)

Alerting Rules

- alert: HighErrorRate
  expr: rate(http_requests_total{status=~"5.."}[5m]) > 0.05
  for: 5m
  annotations:
    summary: "Error rate > 5%"

Runbook Template

## Deployment Runbook

### Prerequisites
- kubectl access, valid kubeconfig
- Registry credentials
- Monitoring access

### Deploy
1. Merge to main → pipeline triggers
2. Monitor pipeline in CI
3. Verify in monitoring dashboard

### Rollback
kubectl rollout undo deployment/app -n production

### Troubleshooting
- Pods not starting: `kubectl logs -f deployment/app`
- High errors: Check app logs and metrics