Memory Management and Eviction

+15 Mana ✨
Redis Memory Optimization(DOCUMENTATION)

Introduction

Redis stores everything in memory, so managing memory is critical. When memory runs out, Redis must decide what to do — reject new writes or evict existing keys. Understanding eviction policies lets you control this behavior.

Key Concepts

  • maxmemory: The maximum amount of memory Redis is allowed to use. Once reached, the eviction policy kicks in.
  • Eviction policy: The algorithm Redis uses to decide which keys to remove when memory is full.
  • Volatile keys: Keys that have an expiration (TTL) set.
  • LRU vs LFU: Least Recently Used evicts keys not accessed recently; Least Frequently Used evicts keys accessed least often overall.

Real World Context

Without a maxmemory limit, Redis grows until the OS kills it (OOM). In production, you always set maxmemory and choose an eviction policy that matches your use case — caches use allkeys-lru, while mixed workloads may use volatile-lru to protect persistent keys.

Deep Dive

Setting Memory Limits

redis
# Set max memory to 256MB
CONFIG SET maxmemory 256mb

# Check current memory usage
INFO memory
# used_memory_human: 45.12M
# maxmemory_human: 256.00M
# maxmemory_policy: noeviction

# Set in redis.conf for persistence
maxmemory 256mb

Eviction Policies

PolicyScopeAlgorithmBest For
noeviction—Reject writesPrimary database
allkeys-lruAll keysLeast Recently UsedGeneral cache
volatile-lruKeys with TTLLeast Recently UsedMixed workload
allkeys-lfuAll keysLeast Frequently UsedFrequency-based cache
volatile-lfuKeys with TTLLeast Frequently UsedPopular item cache
allkeys-randomAll keysRandomWhen all keys are equal
volatile-randomKeys with TTLRandomSimple TTL cache
volatile-ttlKeys with TTLShortest TTL firstTime-sensitive data
redis
# Set eviction policy
CONFIG SET maxmemory-policy allkeys-lru

# Check current policy
CONFIG GET maxmemory-policy

Monitoring Memory

redis
# Detailed memory stats
INFO memory

# Key metrics:
# used_memory: Total bytes allocated
# used_memory_peak: Maximum bytes ever used
# mem_fragmentation_ratio: > 1.5 means fragmentation
# evicted_keys: Number of keys evicted

# Memory usage for a specific key
MEMORY USAGE user:1001
# Returns: 72 (bytes)

# Memory doctor (diagnostics)
MEMORY DOCTOR

Memory Optimization Tips

redis
# Use hashes for small objects (ziplist encoding)
# Instead of:
SET user:1:name "Alice"
SET user:1:email "alice@example.com"

# Use a hash (more memory efficient):
HSET user:1 name "Alice" email "alice@example.com"

# Short keys save memory at scale
SET u:1:n "Alice"    # Saves bytes per key

Common Pitfalls

  1. Not setting maxmemory — Redis grows until the OS kills it with an OOM error. Always set a limit in production.
  2. Using noeviction for caches — With noeviction, Redis returns errors when full instead of evicting old data. Use allkeys-lru or allkeys-lfu for caches.
  3. Ignoring fragmentation — A mem_fragmentation_ratio above 1.5 wastes memory. Restart Redis or enable active defragmentation.

Best Practices

  1. Set maxmemory to 75% of available RAM — Leave room for the OS, background saves (fork), and other processes.
  2. Match policy to workload — Use allkeys-lru for pure caches, volatile-lru when mixing cached and persistent data, and noeviction when every key matters.
  3. Monitor evicted_keys — A rising count means Redis is under memory pressure. Scale up or optimize data structures.

Summary

  • Always set maxmemory in production to prevent OOM kills.
  • Choose an eviction policy that matches your use case (allkeys-lru for caches, noeviction for databases).
  • Monitor memory with INFO memory and watch for fragmentation.
  • Use memory-efficient data structures like hashes for small objects.

Code Examples

bash
# Check memory usage and eviction policy
redis-cli INFO memory | grep -E 'used_memory_human|maxmemory|evicted_keys'
# used_memory_human:45.12M
# maxmemory_human:256.00M
# maxmemory_policy:allkeys-lru
# evicted_keys:0
✓ Completed