Redis Eviction Policy Analyzer

Analyzer ID	Category	Severity	Time To Fix
redis-eviction-policy	✅ Reliability	High	10 minutes

What This Checks

• Queries the live Redis server for the current maxmemory-policy via CONFIG GET
• Validates the eviction policy is appropriate based on how your application uses Redis (cache, queue, session, Horizon)
• Detects when a cache-only Redis server uses noeviction (causes write errors when memory is full)
• Detects when Redis stores persistent data (queues, sessions, Horizon) with an eviction policy that could delete jobs or sessions
• Detects when multiple named connections (e.g. cache and default) resolve to the same host:port — since eviction policy is server-level, a cache-only connection is not flagged if its server also serves persistent data (changing the policy there would be dangerous)
• Checks whether maxmemory is configured (warns if unlimited)
• Monitors memory usage and warns when consumption exceeds 90% of the configured limit
• Only runs in production and staging environments (skipped in CI and local)

Why It Matters

• Data loss: If Redis stores queue jobs or sessions and uses an LRU/LFU eviction policy, pending jobs can be silently deleted when memory pressure occurs
• Write failures: The default noeviction policy returns errors on writes when memory is full -- if Redis is only used for caching, this causes application errors instead of gracefully evicting stale entries
• Silent degradation: Users get logged out unexpectedly when session data is evicted, or background jobs disappear without any error in application logs
• Memory exhaustion: Without a maxmemory limit, Redis can consume all available system memory, causing OOM kills in containerized environments
• Production outages: Memory-related Redis failures cascade to every part of the application that depends on it

How to Fix

Quick Fix (2 minutes)

Check and set the eviction policy directly on your Redis server:

# Check current policy
redis-cli CONFIG GET maxmemory-policy

# For cache-only servers: allow eviction of least-frequently-used keys
redis-cli CONFIG SET maxmemory-policy allkeys-lfu

# For servers storing persistent data (queues, sessions): prevent eviction
redis-cli CONFIG SET maxmemory-policy noeviction

# Set a memory limit (required for eviction policies to work)
redis-cli CONFIG SET maxmemory 2gb

# Persist changes to redis.conf
redis-cli CONFIG REWRITE

Proper Fix (10 minutes)

1: Separate Redis Instances by Workload

The best practice is to use different Redis servers (different host or port) for cache vs persistent data. Different logical databases on the same server (database: 0 vs database: 1) are not sufficient — maxmemory-policy is a server-level setting and applies to all databases on the instance simultaneously.

// config/database.php
'redis' => [
    'client' => env('REDIS_CLIENT', 'phpredis'),

    // Cache instance - use allkeys-lfu eviction
    'cache' => [
        'host' => env('REDIS_CACHE_HOST', '127.0.0.1'),
        'port' => env('REDIS_CACHE_PORT', 6380),
        'database' => env('REDIS_CACHE_DB', 0),
    ],

    // Queue/Session instance - use noeviction
    'persistent' => [
        'host' => env('REDIS_PERSISTENT_HOST', '127.0.0.1'),
        'port' => env('REDIS_PERSISTENT_PORT', 6381),
        'database' => env('REDIS_PERSISTENT_DB', 0),
    ],
],

// config/cache.php
'redis' => [
    'driver' => 'redis',
    'connection' => 'cache',  // Uses cache Redis instance
],

// config/queue.php
'redis' => [
    'driver' => 'redis',
    'connection' => 'persistent',  // Uses persistent Redis instance
    'queue' => 'default',
    'retry_after' => 90,
],

// config/session.php
'driver' => 'redis',
'connection' => 'persistent',  // Uses persistent Redis instance

2: Configure Redis Server Settings

For the cache Redis instance (redis-cache.conf):

maxmemory 2gb
maxmemory-policy allkeys-lfu

For the persistent Redis instance (redis-persistent.conf):

maxmemory 1gb
maxmemory-policy noeviction
appendonly yes

3: Environment-Specific Configuration

# .env (production)
REDIS_CACHE_HOST=redis-cache.internal
REDIS_CACHE_PORT=6379
REDIS_PERSISTENT_HOST=redis-persistent.internal
REDIS_PERSISTENT_PORT=6379

4: Monitor Memory Usage

Set up monitoring to alert before Redis reaches its memory limit:

# Check memory usage
redis-cli INFO memory | grep used_memory_human
redis-cli INFO memory | grep maxmemory_human

# Check eviction stats
redis-cli INFO stats | grep evicted_keys

Eviction Policy Reference

Policy	Behavior	Best For
noeviction	Returns error on write when full	Queues, sessions, persistent data
allkeys-lfu	Evicts least frequently used keys	General-purpose cache (Redis 4+)
allkeys-lru	Evicts least recently used keys	General-purpose cache
volatile-lfu	Evicts LFU keys with TTL set	Mixed cache + persistent on same instance
volatile-lru	Evicts LRU keys with TTL set	Mixed cache + persistent on same instance
volatile-ttl	Evicts keys closest to expiry	Time-sensitive cache entries
allkeys-random	Evicts random keys	Uniform access patterns

ShieldCI Configuration

This analyzer is automatically skipped in CI environments ($runInCI = false) and only runs in production and staging environments.

Why skip in CI and development?

• Eviction policy checks require a live Redis connection with access to the CONFIG GET command
• CI and local Redis instances are typically ephemeral and not configured to reflect production settings
• Eviction policy mismatches only matter when Redis is under real memory pressure

Environment Detection: The analyzer checks your Laravel APP_ENV setting and only runs when it maps to production or staging. Custom environment names can be mapped in config/shieldci.php:

// config/shieldci.php
'environment_mapping' => [
    'production-us' => 'production',
    'production-blue' => 'production',
    'staging-preview' => 'staging',
],

Examples:

• APP_ENV=production → Runs (no mapping needed)
• APP_ENV=production-us → Maps to production → Runs
• APP_ENV=local → Skipped (not production/staging)

Managed Redis (ElastiCache, Upstash, Memorystore, Redis Cloud): Many managed Redis providers restrict the CONFIG GET command via ACL. The analyzer detects this and skips analysis for those connections rather than emitting false positives — no action required on your part.

References

• Redis Eviction Policies
• Redis Configuration - maxmemory
• Laravel Redis Configuration
• Laravel Cache Configuration

Related Analyzers

• Redis Shared Database Analyzer - Detects when cache, queue, and session share the same Redis database
• Cache Prefix Configuration Analyzer - Ensures shared cache backends use unique prefixes
• Cache Status Analyzer - Validates cache connectivity and functionality
• Queue Timeout Configuration Analyzer - Ensures queue timeout and retry_after values are properly configured