redis

Redis Cursor Rules

You are an expert Redis developer. Follow these rules:

Data Structures

• Strings for simple key-value and counters (INCR/DECR are atomic)
• Hashes for objects — HSET user:123 name "Alice" email "alice@..."
• Sorted sets for leaderboards, rate limiters, priority queues
• Lists for queues (LPUSH/RPOP) and activity feeds
• Sets for unique collections and intersection/union operations
• Streams for event sourcing and message queues

Key Design

• Namespace keys with colons: user:123:session, post:456:likes
• Keep keys short but readable — abbreviate namespaces in high-volume scenarios
• Set TTL on every cache key — no key should live forever without reason
• Use SCAN over KEYS — KEYS blocks the single-threaded server

Caching Patterns

• Cache-aside: app checks cache, falls back to DB, writes to cache
• Write-through for data that must stay consistent
• TTL-based invalidation for most use cases. Event-based for critical consistency
• Serialize as JSON or MessagePack — avoid language-specific formats
• Cache stampede prevention: use locks or stale-while-revalidate

Performance

• Pipeline multiple commands to reduce round trips
• Lua scripts for atomic multi-step operations (EVAL/EVALSHA)
• Avoid large values (>100KB) — split into smaller keys
• Use OBJECT ENCODING to verify memory-efficient encodings
• Monitor with INFO stats, SLOWLOG, and CLIENT LIST

Pub/Sub & Streams

• Pub/Sub for fire-and-forget broadcasting — messages are lost if no subscriber
• Streams for persistent messaging with consumer groups
• XACK processed messages. Set MAXLEN to cap stream size
• Use consumer groups for distributed work distribution

Operations

• Redis Sentinel for high availability, Redis Cluster for horizontal scaling
• RDB snapshots for backups. AOF for durability
• maxmemory-policy: allkeys-lru for caches, noeviction for primary data
• Monitor memory with INFO memory — watch fragmentation ratio