1// Making services stateless for horizontal scaling
2 
3// ❌ Stateful — breaks with multiple instances
4class StatefulService {
5  private processingOrders = new Map<string, Order>(); // in-memory
6 
7  async getStatus(orderId: string) {
In-memory state breaks with multiple instances — load balancer routes to different instances
8    return this.processingOrders.get(orderId);  // ❌ Only works on THIS instance
9  }
10}
11 
12// ✅ Stateless — works with any number of instances
13class StatelessService {
14  async startProcessing(orderId: string) {
15    const order = await db.orders.findById(orderId);
16 
17    // Store state in shared Redis — visible to ALL instances
Redis is visible to all instances — the shared source of truth for distributed state
18    await redis.setex(
19      `processing:${orderId}`,
20      3600,
21      JSON.stringify({ status: 'processing', startedAt: Date.now() })
22    );
23  }
24 
Any instance can answer — they all read from the same Redis cluster
25  async getStatus(orderId: string) {
26    const data = await redis.get(`processing:${orderId}`);
27    return data ? JSON.parse(data) : null;
28    // ✅ Any instance can answer — all see the same Redis
29  }
Kubernetes HPA scales pod count automatically based on CPU/memory/custom metrics
30}
31 
32// Kubernetes HorizontalPodAutoscaler
33// apiVersion: autoscaling/v2
34// kind: HorizontalPodAutoscaler
35// spec:
36//   scaleTargetRef:
37//     name: order-service
38//   minReplicas: 2
39//   maxReplicas: 50
40//   metrics:
41//   - type: Resource
42//     resource:
43//       name: cpu
44//       target:
45//         averageUtilization: 70  # scale when avg CPU > 70%