# CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

## 始终使用中文回复

- 在所有沟通和代码注释中，必须使用中文进行交流
- 保持语言的专业性和技术准确性

## mcp 工具使用

- 当涉及到相关库的使用时，应该用 context7 查询相关文档

## Project Overview

FlashSaleSystem is a high-concurrency flash sale (秒杀) system built with Spring Boot and Redis Cluster. The system
implements distributed architecture with anti-overselling mechanisms, rate limiting, and distributed locking using Redis
and Lua scripts.

**Project Completion**: 90%

## Core Architecture

### Technology Stack

- **Backend**: Spring Boot 2.7.6, Java 1.8
- **Database**: MySQL 5.7+ with JPA/Hibernate
- **Cache**: Redis 6.0+ Cluster (6 nodes) with Redisson 3.24.3
- **Frontend**: JSP + JSTL + Bootstrap 5 + jQuery
- **Build Tool**: Maven 3.6+
- **API Documentation**: Knife4j 4.1.0 (Swagger)

### Key Components

- **Controllers**: Handle HTTP requests for different modules (flash sales, cart, orders, admin)
- **Services**: Business logic layer with Redis-based caching and distributed operations
- **Repositories**: JPA data access layer for MySQL operations
- **DTOs**: Data transfer objects for API communication
- **Entities**: JPA entities mapping to database tables

### Package Structure

```
com.org.flashsalesystem/
├── controller/     # REST controllers and web endpoints
├── service/        # Business logic and Redis operations
├── repository/     # JPA repositories for data access
├── entity/         # JPA entities (User, Product, Order, FlashSale)
├── dto/            # Data transfer objects
├── config/         # Configuration classes (Redis, Swagger, Web)
└── util/           # Utility classes and JSP functions
```

### Redis Integration

- **Multiple Modes Supported**: Single node, Cluster (6 nodes), Sentinel
- **Mode Selection**: Automatic based on configuration in application.yml
- **Default Mode**: Single node (localhost:6379) for development
- **Cluster Configuration**: 6-node Redis cluster at 42.192.62.91:7000-7005 via RedissonConfig
- **Data Types**: String (locks, sessions), Hash (user/product info, cart), List (order queues), Set (successful users),
  ZSet (rankings)
- **Lua Scripts**: Atomic operations for flash sales, distributed locks, rate limiting, cart operations
- **Message Queues**: Pub/Sub channels - order:status:change, stock:change, flashsale:result

### Redis Key Prefixes

- `flashsale:` - Flash sale activity data
- `flashsale_stock:` - Real-time stock information
- `flashsale_lock:` - Distributed locks for flash sales
- `flashsale_success:` - Successful user sets
- `user:` - User information cache
- `product:` - Product information cache
- `cart:` - Shopping cart data
- `rate_limit:` - API rate limiting counters

## Common Development Commands

### Build and Run

```bash
# 编译项目
mvn clean compile

# 运行应用（默认单节点Redis）
mvn spring-boot:run

# 运行应用（开发环境，单节点Redis）
mvn spring-boot:run -Dspring.profiles.active=dev

# 运行应用（生产环境，Redis集群）
mvn spring-boot:run -Dspring.profiles.active=cluster

# 打包应用
mvn clean package

# 跳过测试打包
mvn clean package -DskipTests

# 运行打包后的jar（默认配置）
java -jar target/FlashSaleSystem-0.0.1-SNAPSHOT.jar

# 运行打包后的jar（指定profile）
java -jar target/FlashSaleSystem-0.0.1-SNAPSHOT.jar --spring.profiles.active=cluster
```

### Testing

```bash
# 运行所有测试（目前无测试类，需要创建）
mvn test

# 运行特定测试类
mvn test -Dtest=FlashSaleServiceTest
mvn test -Dtest=RedisServiceTest

# 运行集成测试
mvn verify
```

### Database Setup

```bash
# 创建数据库
mysql -u root -p
CREATE DATABASE flash_sale_db CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;

# 导入表结构（JPA会自动创建，但也可以手动导入）
mysql -u root -p flash_sale_db < src/main/resources/sql/schema.sql

# 导入测试数据
mysql -u root -p flash_sale_db < src/main/resources/sql/test-data.sql

# 导入演示用户（包含预设密码）
mysql -u root -p flash_sale_db < src/main/resources/sql/demo-users.sql
```

## Application Configuration

### Key Configuration Files

- `application.yml`: Main application configuration including Redis cluster, database, and custom flash sale settings
- `pom.xml`: Maven dependencies and build configuration
- `src/main/resources/lua/`: Lua scripts for Redis atomic operations

### Important Configuration Sections

- **Redis Mode**: Supports Single node (default), Cluster (6 nodes), Sentinel modes
- **Redis Single Node**: localhost:6379 (default for development)
- **Redis Cluster**: 6-node cluster with password authentication (6HU3cw1drNjfQ0zo1Uyx)
- **Flash Sale Settings**: Rate limiting (10 requests/minute), max quantity per user (1), stock preload (30min advance)
- **Cache TTL**: User info (30min), product info (60min), flash sale data (10min)
- **Database**: HikariCP connection pool (max 20, min 5)
- **Cart Settings**: Expire after 7 days, max 20 items

### Redis Mode Switching

1. **Single Node Mode** (Default):
  - Edit `application.yml` and ensure cluster configuration is commented out
  - Or use profile: `mvn spring-boot:run -Dspring.profiles.active=dev`

2. **Cluster Mode**:
  - Use profile: `mvn spring-boot:run -Dspring.profiles.active=cluster`
  - Or uncomment cluster configuration in `application.yml`

3. **Environment Variable Override**:
   ```bash
   export SPRING_REDIS_HOST=your-redis-host
   export SPRING_REDIS_CLUSTER_NODES=node1:7000,node2:7001
   ```

## Development Guidelines

### Service Layer Patterns

- All Redis operations use Redisson for cluster support
- Critical operations (stock deduction, cart updates) use Lua scripts for atomicity
- Distributed locks prevent overselling using Redisson's RLock with timeout
- Rate limiting implemented via sliding window algorithm in Lua
- Use RedisPipelineService for batch operations to improve performance

### Key Services

- **FlashSaleService**: Core flash sale logic with distributed locking and Lua scripts
- **RedisService**: Generic Redis operations wrapper with TTL support
- **DistributedLockService**: Distributed locking abstraction with auto-renewal
- **RateLimitService**: API rate limiting using Redis sliding window
- **CartService**: Shopping cart operations with Redis Hash storage
- **RedisPipelineService**: Batch Redis operations using pipeline
- **MessageListenerService**: Redis Pub/Sub message handling

### Flash Sale Process Flow

1. **Stock Preload**: FlashSaleService.preloadStock() loads stock to Redis 30min before
2. **Rate Limit Check**: RateLimitService.isAllowed() checks request frequency
3. **Distributed Lock**: DistributedLockService.tryLock() prevents concurrent access
4. **Stock Check**: Lua script atomically checks and deducts stock
5. **Order Creation**: Creates order in database and publishes result
6. **Cache Update**: Updates user success set and publishes notifications

### Testing Approach

- Unit tests should be created for core services (FlashSaleService, RedisService, CartService)
- Integration tests should verify Redis cluster connectivity and database operations
- Load testing recommended using JMeter for flash sale scenarios
- Mock Redis operations in unit tests using Mockito

## Lua Script Usage

The system includes 5 Lua scripts in `src/main/resources/lua/`:

- `flashsale.lua`: Atomic stock deduction with overselling prevention (returns: remaining stock or negative error codes)
- `distributed_lock.lua`: Distributed lock acquisition with expiry
- `unlock.lua`: Safe distributed lock release with owner verification
- `rate_limit.lua`: Sliding window rate limiting (returns: 1 if allowed, 0 if blocked)
- `cart_operation.lua`: Atomic shopping cart operations (add/update/remove items)

### Lua Script Error Codes

- `-1`: Key not exists or invalid
- `-2`: Insufficient stock
- `-3`: Invalid parameter

## API Documentation

- **Swagger UI**: http://localhost:8080/doc.html (Knife4j enhanced UI)
- **API Docs**: http://localhost:8080/v3/api-docs
- **Test Endpoints**: TestController provides Redis connection and data verification endpoints

## Monitoring and Health

- **Actuator Endpoints**: /actuator/health, /actuator/metrics, /actuator/prometheus
- **Log Files**: logs/flash-sale-system.log with detailed Redis and SQL logging
- **Debug Level**: Enabled for com.org.flashsalesystem, org.springframework.data.redis, org.hibernate.SQL
- **Metrics Export**: Prometheus metrics enabled for monitoring

## Security Considerations

- **Redis Authentication**: Cluster uses password authentication
- **Database Connection**: Uses HikariCP with SSL disabled (enable for production)
- **Password Encoding**: BCrypt with strength 10 for user passwords
- **Rate Limiting**: Built-in rate limiting to prevent abuse
- **Session Management**: Redis-based session storage with 30min timeout

## Performance Optimization

- **Connection Pooling**: HikariCP (20 max) for DB, Jedis pool (20 max) for Redis
- **Pipeline Operations**: Use RedisPipelineService for batch operations
- **Cache Strategy**: Multi-level caching with appropriate TTL
- **Lua Scripts**: Reduce network round trips for atomic operations
- **Stock Preload**: Warm up cache before flash sale starts

## Common Issues and Solutions

1. **Redis Connection Failed**: Check cluster nodes accessibility and password
2. **Stock Overselling**: Ensure Lua scripts are loaded and distributed locks work
3. **High Latency**: Use pipeline for batch operations, check network latency
4. **Memory Issues**: Monitor Redis memory usage, adjust cache TTL if needed