Redis：优雅键值设计与BigKey处理指南

1. Elegant Key Structure

Redis keys should follow best practices, including a basic format of [business name]:[data name]:[data/data id], with a maximum length of 44 bytes, and no special characters for benefits like readability, conflict avoidance, easy management, and memory efficiency. For example, user information for a login service might be formatted as login:user:12345. Keys shorter than 44 bytes use the more memory-efficient embstr encoding.

2. Cautious Use of BigKeys

BigKeys can be identified by their size or member count, such as a 5 MB String key or a ZSET with 10,000 members. It is recommended to keep single key values under 10 KB and collections under 1000 elements, preferably not exceeding 5000. Redis introduced the MEMORY command in version 4.0 for size assessments, but its use is generally not recommended due to high CPU usage.

2.1. Dangers of BigKeys

BigKeys can cause network blockages, data skew, Redis blocking, and high CPU pressure, leading to performance degradation and imbalance in memory resources among data shards.

3. How to Discover BigKeys

To find BigKeys, use the redis-cli --bigkeys option for a summary of keys, or programmatically scan all keys and assess sizes with commands like strlen and hlen. Using third-party tools like Redis-Rdb-Tools for RDB file analysis can provide a comprehensive memory usage analysis.

4. How to Delete BigKeys

Deleting BigKeys can be time-consuming and block the main thread of Redis. In versions before 3.0, it's suggested to remove elements of a collection one by one before deleting the entire BigKey. The DEL command can remove keys but may impact performance due to its synchronous nature. Redis 4.0 introduced the unlink command for asynchronous deletion.