Performance Issues

Solutions for latency and throughput problems.

Problem: High Latency

Symptoms

• Redis operations taking >10ms
• Increased p99 latency
• Slow API responses

Diagnosis

# Check Redis latency
redis-cli --latency

# Check slow log
redis-cli SLOWLOG GET 10

# Check network latency
redis-cli DEBUG SLEEP 0 && time redis-cli PING

# Operation latency p99
histogram_quantile(0.99, rate(redisx_operation_duration_seconds_bucket[5m]))

Solutions

Cause	Solution
Network latency	Move Redis closer to app
Large values	Compress or split data
Expensive commands	Avoid KEYS, use SCAN
Connection pool exhausted	Increase pool size
Redis overloaded	Scale up or out

---

Problem: Connection Pool Exhaustion

Symptoms

• "Cannot acquire connection" errors
• Requests timing out waiting for connection
• Spiky latency under load

Diagnosis

# Pool utilization
redisx_pool_active_connections / redisx_pool_max_connections

# Pending requests
redisx_pool_pending_requests

Solutions

// Increase pool size
RedisModule.forRoot({
  clients: {
    host: 'localhost',
    pool: {
      min: 10,
      max: 50,  // Increase from default
    },
  },
})

Cause	Solution
Pool too small	Increase pool.max
Connections not released	Check for connection leaks
Slow commands	Optimize or add timeout

---

Problem: Memory Pressure

Symptoms

• Redis memory near max
• Evictions increasing
• Slow operations

Diagnosis

# Memory info
redis-cli INFO memory

# Find big keys
redis-cli --bigkeys

# Memory usage by key
redis-cli MEMORY USAGE "cache:big-key"

Solutions

Cause	Solution
TTL too long	Reduce TTL
Large values	Compress data
Too many keys	Reduce cardinality
No eviction policy	Set maxmemory-policy

# Set eviction policy
redis-cli CONFIG SET maxmemory-policy allkeys-lru

---

Problem: Slow Commands

Symptoms

• Specific operations are slow
• SLOWLOG shows patterns
• High CPU on Redis

Diagnosis

# Check slow commands
redis-cli SLOWLOG GET 20

Production Risk

redis-cli MONITOR impacts performance significantly. Never use in production.

Solutions

Slow Command	Alternative
KEYS *	SCAN with cursor
SMEMBERS large set	SSCAN
HGETALL large hash	HSCAN or specific HGET
Lua with many keys	Break into smaller operations

Correct

// Non-blocking with SCAN
const keys = [];
let cursor = '0';
do {
  const [newCursor, batch] = await redis.scan(cursor, 'MATCH', 'user:*', 'COUNT', 100);
  cursor = newCursor;
  keys.push(...batch);
} while (cursor !== '0');

Wrong

// Blocks Redis - never use in production
await redis.keys('user:*');

---

Problem: Throughput Bottleneck

Symptoms

• Can't scale beyond certain RPS
• Redis CPU at 100%
• Adding app instances doesn't help

Diagnosis

# Check Redis stats
redis-cli INFO stats

# Check CPU
redis-cli INFO cpu

Solutions

Cause	Solution
Single-threaded Redis	Use Redis Cluster
CPU-intensive commands	Optimize commands
Too many round trips	Use pipelining

Correct

// Single round trip with pipeline
const pipeline = redis.pipeline();
pipeline.set('key1', 'val1');
pipeline.set('key2', 'val2');
pipeline.set('key3', 'val3');
await pipeline.exec();

Wrong

// Multiple round trips - inefficient
await redis.set('key1', 'val1');
await redis.set('key2', 'val2');
await redis.set('key3', 'val3');

---

Problem: Cold Start Latency

Symptoms

• First requests after deploy are slow
• Cache miss spike after restart
• Database overwhelmed on deploy

Solutions

// Warm cache on startup
@Injectable()
export class CacheWarmer implements OnModuleInit {
  async onModuleInit() {
    // Pre-populate critical cache
    await this.warmPopularProducts();
    await this.warmFeatureFlags();
  }

  private async warmPopularProducts() {
    const products = await this.db.getPopularProducts(100);
    for (const product of products) {
      await this.cache.set(`product:${product.id}`, product, { ttl: 3600 });
    }
  }
}

Performance Checklist

Before Going Live

Verify these items for optimal Redis performance:

Item	Status
Connection pool sized appropriately
TTLs not too long or short
No KEYS commands in production
Large values compressed
Pipelining used where possible
Redis topology matches workload
Monitoring in place

Next Steps

• Debugging — Debugging tools
• Tuning — Performance optimization