When testing in production or implementing progressive delivery:

IMPLEMENT feature flags for progressive rollout (1% → 10% → 50% → 100%)
DEPLOY with canary releases (compare metrics before full rollout)
MONITOR with synthetic tests (proactive) + RUM (reactive)
INJECT failures with chaos engineering (build resilience)
ANALYZE production data to improve pre-production testing

Quick Shift-Right Techniques:

Feature flags → Control who sees what, instant rollback
Canary deployment → 5% traffic, compare error rates
Synthetic monitoring → Simulate users 24/7, catch issues before users
Chaos engineering → Netflix-style failure injection
RUM (Real User Monitoring) → Actual user experience data

Critical Success Factors:

Production is the ultimate test environment
Ship fast with safety nets, not slow with certainty
Use production data to improve shift-left testing

Quick Reference Card

When to Use

Progressive feature rollouts
Production reliability validation
Performance monitoring at scale
Learning from real user behavior

Shift-Right Techniques

| Feature Flags | Controlled rollout | Every feature

| Canary | Compare new vs old | Every deployment

| Synthetic Monitoring | Proactive detection | 24/7

| RUM | Real user metrics | Always on

| Chaos Engineering | Resilience validation | Regularly

| A/B Testing | User behavior validation | Feature decisions

Progressive Rollout Pattern

1% → 10% → 25% → 50% → 100%
↓      ↓      ↓      ↓
Check  Check  Check  Monitor

Key Metrics to Monitor

| Error rate | < 0.1% | > 1%

| p95 latency | < 200ms | > 500ms

| Availability | 99.9% | < 99.5%

| Apdex | > 0.95 | < 0.8

Feature Flags

// Progressive rollout with LaunchDarkly/Unleash pattern
const newCheckout = featureFlags.isEnabled('new-checkout', {
  userId: user.id,
  percentage: 10,  // 10% of users
  allowlist: ['beta-testers']
});

if (newCheckout) {
  return <NewCheckoutFlow />;
} else {
  return <LegacyCheckoutFlow />;
}

// Instant rollback on issues
await featureFlags.disable('new-checkout');

Canary Deployment

# Flagger canary config
apiVersion: flagger.app/v1beta1
kind: Canary
spec:
  targetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: checkout-service
  progressDeadlineSeconds: 60
  analysis:
    interval: 1m
    threshold: 5      # Max failed checks
    maxWeight: 50     # Max traffic to canary
    stepWeight: 10    # Increment per interval
    metrics:
      - name: request-success-rate
        threshold: 99
      - name: request-duration
        threshold: 500

Synthetic Monitoring

// Continuous production validation
await Task("Synthetic Tests", {
  endpoints: [
    { path: '/health', expected: 200, interval: '30s' },
    { path: '/api/products', expected: 200, interval: '1m' },
    { path: '/checkout', flow: 'full-purchase', interval: '5m' }
  ],
  locations: ['us-east', 'eu-west', 'ap-south'],
  alertOn: {
    statusCode: '!= 200',
    latency: '> 500ms',
    contentMismatch: true
  }
}, "qe-production-intelligence");

Chaos Engineering

// Controlled failure injection
await Task("Chaos Experiment", {
  hypothesis: 'System handles database latency gracefully',
  steadyState: {
    metric: 'error_rate',
    expected: '< 0.1%'
  },
  experiment: {
    type: 'network-latency',
    target: 'database',
    delay: '500ms',
    duration: '5m'
  },
  rollback: {
    automatic: true,
    trigger: 'error_rate > 5%'
  }
}, "qe-chaos-engineer");

Production → Pre-Production Feedback Loop

// Convert production incidents to regression tests
await Task("Incident Replay", {
  incident: {
    id: 'INC-2024-001',
    type: 'performance-degradation',
    conditions: { concurrent_users: 500, cart_items: 10 }
  },
  generateTests: true,
  addToRegression: true
}, "qe-production-intelligence");

// Output: New test added to prevent recurrence

Agent Coordination Hints

Memory Namespace

aqe/shift-right/
├── canary-results/*      - Canary deployment metrics
├── synthetic-tests/*     - Monitoring configurations
├── chaos-experiments/*   - Experiment results
├── production-insights/* - Issues → test conversions
└── rum-analysis/*        - Real user data patterns

Fleet Coordination

const shiftRightFleet = await FleetManager.coordinate({
  strategy: 'shift-right-testing',
  agents: [
    'qe-production-intelligence',  // RUM, incident replay
    'qe-chaos-engineer',           // Resilience testing
    'qe-performance-tester',       // Synthetic monitoring
    'qe-quality-analyzer'          // Metrics analysis
  ],
  topology: 'mesh'
});

shift-left-testing - Pre-production testing
chaos-engineering-resilience - Failure injection deep dive
performance-testing - Load testing
agentic-quality-engineering - Agent coordination

Remember

Production is the ultimate test environment. Feature flags enable instant rollback. Canary catches issues before 100% rollout. Synthetic monitoring detects problems before users. Chaos engineering builds resilience. RUM shows real user experience.

With Agents: Agents monitor production, replay incidents as tests, run chaos experiments, and convert production insights to pre-production tests. Use agents to maintain continuous production quality.

安装