Flash Sale with Limited Tickets
Problem Statements Limited inventory: Exactly 200 tickets, no overselling. One ticket per user: Strictly enforce (assuming logged-in users only). Handle high concurrency: 20k+ users rushing at sale start, avoid server overload/high peak load. Minimize database stress: Avoid hammering the persistent DB during the peak. Typical Solutions Fitting Your Node.js Setup To meet your requirements: Use Redis as the primary hotspot for inventory and user checks (in-memory, distributed via clustering). Pre-load remaining tickets as a counter (e.g., atomic DECR). Use a Set for sold users (check/add atomically). Best: Redis Lua script (or Redis function in v7+) for atomicity: check stock >0, check user not bought, decrement stock, add user. Enforce one per user via user ID in a Redis Set. Handle concurrency: Virtual queue (e.g., Redis list or separate service like RabbitMQ/Kafka) or waiting room to throttle ingress. Offload DB: Successful attempts go to a message queue (e.g., Kafka/RabbitMQ) for async persistence; failures reject immediately. Node.js scaling: Cluster with PM2, share Redis, use worker threads for non-blocking I/O. This pattern prevents overselling reliably while keeping peak DB hits near zero. ...