12. Common Issues & Troubleshooting

Q1: Connection Refused

# Error message
Could not connect to Redis at 127.0.0.1:6379: Connection refused

# Troubleshooting steps
# 1. Check if Redis is running
sudo systemctl status redis-server
ps aux | grep redis

# 2. Check if port is listening
ss -tlnp | grep 6379

# 3. Check firewall
sudo ufw allow 6379
sudo iptables -L -n | grep 6379

# 4. Check bind configuration
CONFIG GET bind

redis-server.service - Advanced key-value store
   Active: active (running) since Sat 2024-01-01 10:00:00 CST
LISTEN  0  128  127.0.0.1:6379  0.0.0.0:*  users:(("redis-server",pid=1234,fd=6))
1) "bind"
2) "127.0.0.1"

Q2: Out of Memory (OOM)

# Error message
OOM command not allowed when used memory > 'maxmemory'.

# Troubleshooting
INFO memory
# Look at: used_memory_human, maxmemory_human, mem_fragmentation_ratio

# Solutions
# 1. Increase maxmemory
CONFIG SET maxmemory 8gb

# 2. Set a reasonable eviction policy
CONFIG SET maxmemory-policy allkeys-lru

# 3. Clean up expired data
SCAN 0 MATCH "temp:*" COUNT 1000

# 4. Check fragmentation ratio (if > 1.5, consider cleanup)
CONFIG SET activedefrag yes

# Memory
used_memory:4294967296
used_memory_human:4.00G
maxmemory:4294967296
maxmemory_human:4.00G
mem_fragmentation_ratio:1.25
OK
OK

Q3: Master-Replica Sync Delay

# Check replication offset difference
INFO replication

# Key fields to watch:
# master_repl_offset vs slave_repl_offset
# Larger difference = more delay

# Solutions
# 1. Ensure sufficient network bandwidth
# 2. Avoid big key writes (writing a List with 1 million elements at once)
# 3. Increase repl_backlog_size
CONFIG SET repl-backlog-size 256mb

# Replication
role:master
connected_slaves:1
slave0:ip=192.168.1.102,port=6379,state=online,offset=1234567,lag=5
master_repl_offset:1234567
OK

Q4: Cache Penetration / Breakdown / Avalanche

Problem	Description	Solution
Cache Penetration	Querying data that doesn't exist — both cache and database return nothing	Bloom filter / Cache null values (with short TTL)
Cache Breakdown	Hot key expires instantly, flooding requests hit the database	Mutex lock (SETNX) / Logical expiration / Never expire
Cache Avalanche	Many keys expire simultaneously, all requests hit the database	Add random value to TTL / Multi-level cache / Circuit breaker & degradation

# Prevent cache avalanche: add random offset to expiration
SET product:1001 "{json_data}" EX 3540  # 3600 - random(0, 120)

# Prevent cache breakdown: distributed lock
# Only one thread queries database and writes to cache
SET lock:product:1001 "1" NX EX 5