Aller au contenu principal

Release Engineering

1 - Principes​

1.1 Objectifs​

release_engineering_goals:
  velocity:
    - Deploy frequently
    - Reduce batch size
    - Shorten feedback loop
    
  safety:
    - Minimize risk
    - Easy rollbacks
    - Controlled blast radius
    
  consistency:
    - Reproducible builds
    - Automated pipelines
    - Standardized processes

1.2 Release Philosophy​

2 - Deployment Strategies​

2.1 Comparatif​

StrategyRiskRollbackComplexity
Big BangHighHardLow
RollingMediumMediumMedium
Blue-GreenLowEasyMedium
CanaryVery LowEasyHigh
Feature FlagsVery LowInstantHigh

2.2 Rolling Deployment​

# Kubernetes Rolling Update
apiVersion: apps/v1
kind: Deployment
spec:
  replicas: 10
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxSurge: 2        # +2 pods max
      maxUnavailable: 1  # -1 pod max

2.3 Blue-Green Deployment​

# Switch via Kubernetes Service
apiVersion: v1
kind: Service
metadata:
  name: api
spec:
  selector:
    app: api
    version: green  # Switch from blue to green

2.4 Canary Deployment​

# Argo Rollouts Canary
apiVersion: argoproj.io/v1alpha1
kind: Rollout
metadata:
  name: api
spec:
  replicas: 10
  strategy:
    canary:
      steps:
        - setWeight: 10
        - pause: { duration: 5m }
        - setWeight: 25
        - pause: { duration: 10m }
        - setWeight: 50
        - pause: { duration: 10m }
        - setWeight: 100
      canaryService: api-canary
      stableService: api-stable

3 - Feature Flags​

3.1 Types​

feature_flag_types:
  release_toggle:
    description: "Enable feature for gradual rollout"
    lifetime: "Short-term"
    
  experiment_toggle:
    description: "A/B testing"
    lifetime: "Medium-term"
    
  ops_toggle:
    description: "Circuit breaker, kill switch"
    lifetime: "Permanent"
    
  permission_toggle:
    description: "Feature gating by user"
    lifetime: "Long-term"

3.2 Implementation​

// Feature flag SDK
import { FeatureFlags } from '@company/feature-flags';

const flags = new FeatureFlags({
  apiKey: process.env.FF_API_KEY,
  environment: process.env.NODE_ENV,
});

// Usage
app.get('/api/checkout', async (req, res) => {
  const user = req.user;
  
  if (await flags.isEnabled('new-checkout', user)) {
    // New checkout flow
    return newCheckout(req, res);
  }
  
  // Old checkout flow
  return oldCheckout(req, res);
});

3.3 Gradual Rollout​

feature_rollout:
  feature: new-checkout
  
  stages:
    - name: internal
      percentage: 0
      users: ["@company.com"]
      
    - name: beta
      percentage: 5
      criteria: "beta_users = true"
      
    - name: early_adopters
      percentage: 25
      
    - name: general_availability
      percentage: 100

4 - Rollback Strategies​

4.1 Types de rollback​

rollback_types:
  version_rollback:
    method: "Deploy previous version"
    time: "Minutes"
    use_case: "Code bugs"
    
  config_rollback:
    method: "Revert configuration"
    time: "Seconds"
    use_case: "Config errors"
    
  feature_flag:
    method: "Disable flag"
    time: "Instant"
    use_case: "Feature issues"
    
  traffic_shift:
    method: "Shift traffic to stable"
    time: "Seconds"
    use_case: "Canary issues"

4.2 Automatic Rollback​

# Argo Rollouts with auto-rollback
apiVersion: argoproj.io/v1alpha1
kind: Rollout
spec:
  strategy:
    canary:
      analysis:
        templates:
          - templateName: success-rate
        args:
          - name: service-name
            value: api
      steps:
        - setWeight: 10
        - pause: { duration: 5m }
        - analysis:
            templates:
              - templateName: success-rate
---
apiVersion: argoproj.io/v1alpha1
kind: AnalysisTemplate
metadata:
  name: success-rate
spec:
  metrics:
    - name: success-rate
      interval: 1m
      successCondition: result[0] > 0.95
      provider:
        prometheus:
          address: http://prometheus:9090
          query: |
            sum(rate(http_requests_total{status=~"2.."}[5m]))
            /
            sum(rate(http_requests_total[5m]))

4.3 Rollback Checklist​

rollback_checklist:
  before:
    - [ ] Confirm rollback is needed
    - [ ] Notify stakeholders
    - [ ] Check rollback target version
    
  during:
    - [ ] Execute rollback
    - [ ] Monitor metrics
    - [ ] Verify user impact
    
  after:
    - [ ] Confirm rollback success
    - [ ] Update status page
    - [ ] Schedule post-mortem
    - [ ] Track root cause

5 - Progressive Delivery​

5.1 Pipeline​

5.2 Automated Promotion​

progressive_delivery:
  stages:
    staging:
      auto_promote: true
      conditions:
        - tests_pass
        - no_critical_vulnerabilities
        
    canary:
      percentage: 5
      duration: 30m
      auto_promote: true
      conditions:
        - error_rate < 1%
        - latency_p99 < 200ms
        
    expand:
      percentage: 25
      duration: 1h
      auto_promote: true
      conditions:
        - error_rate < 0.5%
        - latency_p99 < 200ms
        - no_increase_in_errors
        
    production:
      percentage: 100
      manual_approval: false

6 - Release Documentation​

6.1 Changelog​

# Changelog

## [2.5.0] - 2024-01-15

### Added
- New checkout flow with Apple Pay support
- Real-time inventory tracking

### Changed
- Improved cart performance (30% faster)
- Updated payment SDK to v3.2

### Fixed
- Fixed currency conversion bug (#1234)
- Resolved timeout issues on mobile

### Security
- Updated dependencies for CVE-2024-1234

### Deprecated
- Legacy checkout API (removal in v3.0)

6.2 Release Notes​

release_notes:
  version: "2.5.0"
  date: "2024-01-15"
  
  highlights:
    - "Apple Pay support"
    - "30% faster cart"
    
  user_impact:
    - "Customers can now use Apple Pay"
    - "Faster shopping experience"
    
  technical_changes:
    - "New payment integration"
    - "Cart caching improvements"
    
  known_issues:
    - "Apple Pay not available in some regions"
    
  rollback_instructions:
    - "kubectl rollout undo deployment/checkout"

Résumé​

Dans ce chapitre, nous avons appris :

  • Les principes Release Engineering
  • Les stratégies de déploiement
  • Les Feature Flags
  • Les stratégies de Rollback
  • La Progressive Delivery
  • La documentation des releases

Prochaine étape​

Dans le prochain chapitre, nous verrons les Bonnes pratiques SRE.

→ Chapitre suivant : Bonnes pratiques

← Retour à la table des matières