redis-best-practices

Use Strings for simple caching and counters, Hashes for multi-field objects, Lists for queues and activity feeds, Sets for unique collections and membership checks, Sorted Sets for leaderboards and time-series, and Streams for event logs with consumer groups.

Use a Lua script that atomically increments a counter key and sets an expiration on first increment, returning 0 when the limit is exceeded. Running it via EVAL guarantees atomicity, preventing race conditions that occur with separate INCR and EXPIRE commands.

Use Sentinel for automatic failover with a single primary and read replicas. Use Cluster when you need horizontal scaling, since it shards data across nodes automatically and supports hash tags to keep related keys in the same slot.

KEYS scans the entire keyspace and blocks the server on large datasets. Use SCAN with MATCH and COUNT options instead, which iterates incrementally without blocking other clients.

Set maxmemory to a defined limit and use the allkeys-lru policy, which evicts the least recently used keys when memory is full. Always set TTLs on cache keys and add jitter to expiration times to prevent thundering herd on simultaneous expiry.