Redis Cheatsheet
A production reference for Redis keyspace commands, fundamental data structures, transactions, publishing/subscribing, and client connection setups.
1. General Keyspace Commands
SET user:1 "Jules" # Set string key with value
GET user:1 # Retrieve string key
EXISTS user:1 # Check key existence (returns 1 or 0)
DEL user:1 # Delete a key
EXPIRE user:1 3600 # Set key expiration in seconds (1 hour)
TTL user:1 # Check remaining Time-To-Live in seconds
KEYS * # List all keys (⚠️ Warning: DO NOT use in production!)
SCAN 0 MATCH user:* COUNT 10 # Safe, non-blocking keyspace cursor iteration
FLUSHDB # Delete all keys in current database
FLUSHALL # Delete all keys across all Redis databases
2. Core Data Structures
Lists (Ordered sequences)
LPUSH tasks "Deploy app" # Push item to the left (front)
RPUSH tasks "Write tests" # Push item to the right (back)
LPOP tasks # Pop item from the left
RPOP tasks # Pop item from the right
LRANGE tasks 0 -1 # List all items (0 to end index)
LLEN tasks # Get total list length
Sets (Unordered, unique collections)
SADD tags "python" "redis" # Add elements to the set
SREM tags "redis" # Remove element from the set
SISMEMBER tags "python" # Check membership (returns 1 or 0)
SMEMBERS tags # List all elements in set
SINTER tags1 tags2 # Find intersection of two sets
SUNION tags1 tags2 # Find union of two sets
Hashes (Field-value objects)
HSET user:101 name "Rao" email "rao@ex.com" # Set multiple hash fields
HGET user:101 name # Get specific field
HGETALL user:101 # Get all fields and values
HDEL user:101 email # Delete field
HEXISTS user:101 name # Check field existence
Sorted Sets (Sets ordered by score)
ZADD leaderboard 100 "PlayerA" # Add element with score 100
ZADD leaderboard 150 "PlayerB" # Add element with score 150
ZRANGE leaderboard 0 -1 WITHSCORES # Get all elements sorted lowest to highest
ZREVRANGE leaderboard 0 -1 WITHSCORES # Get elements sorted highest to lowest
ZREM leaderboard "PlayerA" # Remove element
3. Transactions & Pipelining
Redis transactions ensure command execution occurs in a single sequential block.
MULTI # Start transaction block
SET balance:1 500 # Queued
INCRBY balance:1 100 # Queued
EXEC # Execute all queued commands atomically
# Alternatively, discard transaction:
DISCARD # Cancel transaction, discard queued commands
Optimistic Locking via WATCH
WATCH balance:1 # Watch a key for modifications
# Run GET, check value in client...
MULTI
DECRBY balance:1 50
EXEC # Fails (returns null) if balance:1 changed after WATCH
4. Pub/Sub (Publish & Subscribe)
Real-time message broadcasting channel model.
SUBSCRIBE updates:alerts # Client A: Subscribe to channel
PUBLISH updates:alerts "CPU > 90%" # Client B: Broadcast message to channel
PSUBSCRIBE updates:* # Subscribe to all channels matching pattern
5. Python Integration Example
import redis
# Initialize client
r = redis.Redis(host='localhost', port=6379, db=0, decode_responses=True)
# Basic commands
r.set('system:status', 'active', ex=300) # set with 5min TTL
status = r.get('system:status')
print(f"Status: {status}")
# Pipe multiple operations to reduce network RTT latency
pipe = r.pipeline()
pipe.set('temp:1', 'a')
pipe.set('temp:2', 'b')
pipe.execute()
Best Practices & Production Standards
- Reuse Connection Pools: Avoid repeatedly establishing connections; instantiate global connection pools.
- Choose Appropriate Eviction Policies: Set
maxmemory-policy(likeallkeys-lruorvolatile-lru) to guarantee predictable memory behavior. - Pipeline Bulk Queries: Group sequential commands into a single
PIPELINEbatch to eliminate round-trip network times.
Common Mistakes & Antipatterns
- Blocking O(N) Operations on Production: Running commands like
KEYS *,SMEMBERS, orHGETALLon large datasets, blocking Redis’s single-threaded event loop. - Failing to Set Expiries (TTL): Forgetting to apply Time-To-Live parameters to transient cache keys, leading to memory exhaustion.
- Misconfiguring RDB/AOF Persistence: Enabling aggressive disk flushing configurations, causing write-performance degradation.
Troubleshooting & Debugging Guide
- Locating CPU Bottlenecks: Run
SLOWLOG GETto identify queries taking longer than configured thresholds. - Mitigating Cache Stampedes: Implement probabilistic early expiration or mutual exclusion locking (mutex) to prevent concurrent backend queries when cache keys expire.
Core Interview Questions & Answers
- Q: Why is Redis single-threaded, and how does it achieve high performance?
- A: Redis is single-threaded to avoid context switching, race conditions, and thread locks. It achieves high throughput because it operates completely in-memory and uses non-blocking asynchronous I/O multiplexing (via epoll or kqueue) to handle thousands of connections concurrently.
- Q: Compare RDB snapshots and AOF persistence in Redis.
- A: RDB (Redis Database) takes compact point-in-time binary snapshots of memory at intervals (fast recovery, low performance overhead, but risks losing recent writes). AOF (Append-Only File) logs every write operation sequentially (safe durability, but files are larger and slow down recovery).
Technical Architecture Diagram
graph LR
Client[Client App] --> Multiplexer[I/O Multiplexer epoll]
Multiplexer --> Loop[Single Threaded Event Loop]
Loop --> Cache[(In-Memory Key Value Store)]
Loop -.-> Disk[AOF / RDB Background Write]