redis-best-practices

Redis 最佳实践

核心原则

• 使用 Redis 进行缓存、会话存储、实时分析和消息队列
• 根据使用场景选择合适的数据结构
• 制定恰当的键命名规范和过期策略
• 根据高可用性和持久化需求进行设计
• 监控内存使用情况并优化性能

键命名规范

• 使用冒号分隔命名空间
• 在键名中包含对象类型和标识符
• 键名应简短且具有描述性
• 在整个应用中使用一致的命名模式

# Good key naming examples
user:1234:profile
user:1234:sessions
order:5678:items
cache:api:products:list
queue:email:pending
session:abc123def456
rate_limit:api:user:1234

数据结构

字符串

• 用于简单的键值存储、计数器和缓存
• 可考虑使用 MGET/MSET 进行批量操作

# Simple caching
SET cache:user:1234 '{"name":"John","email":"[email protected]"}' EX 3600

# Counters
INCR stats:pageviews:homepage
INCRBY stats:downloads:file123 5

# Atomic operations
SETNX lock:resource:456 "owner:abc" EX 30

哈希

• 用于包含多个字段的对象
• 比使用多个字符串键更节省内存
• 支持部分更新

# Store user profile
HSET user:1234 name "John Doe" email "[email protected]" created_at "2024-01-15"

# Get specific fields
HGET user:1234 email
HMGET user:1234 name email

# Increment numeric fields
HINCRBY user:1234 login_count 1

# Get all fields
HGETALL user:1234

列表

• 用于队列、最近项目和活动动态
• 可考虑为队列消费者使用阻塞操作

# Message queue
LPUSH queue:emails '{"to":"[email protected]","subject":"Welcome"}'
RPOP queue:emails

# Blocking pop for workers
BRPOP queue:emails 30

# Recent activity (keep last 100)
LPUSH user:1234:activity "viewed product 567"
LTRIM user:1234:activity 0 99

# Get recent items
LRANGE user:1234:activity 0 9

集合

• 用于唯一值集合、标签和关系
• 支持集合运算（并集、交集、差集）

# User tags/interests
SADD user:1234:interests "technology" "music" "travel"

# Check membership
SISMEMBER user:1234:interests "music"

# Find common interests
SINTER user:1234:interests user:5678:interests

# Online users tracking
SADD online:users "user:1234"
SREM online:users "user:1234"
SMEMBERS online:users

有序集合

• 用于排行榜、优先级队列和时间序列数据
• 元素按分数排序

# Leaderboard
ZADD leaderboard:game1 1500 "player:123" 2000 "player:456" 1800 "player:789"

# Get top 10
ZREVRANGE leaderboard:game1 0 9 WITHSCORES

# Get player rank
ZREVRANK leaderboard:game1 "player:123"

# Time-based data (score = timestamp)
ZADD events:user:1234 1705329600 "login" 1705330000 "purchase"

# Get events in time range
ZRANGEBYSCORE events:user:1234 1705329600 1705333200

流

• 用于事件流和日志数据
• 支持消费者组，以便进行分布式处理

# Add events to stream
XADD events:orders * customer_id 1234 product_id 567 amount 99.99

# Read from stream
XREAD COUNT 10 STREAMS events:orders 0

# Consumer groups
XGROUP CREATE events:orders order-processors $ MKSTREAM
XREADGROUP GROUP order-processors worker1 COUNT 10 STREAMS events:orders >

# Acknowledge processed messages
XACK events:orders order-processors 1234567890-0

缓存模式

Cache-Aside 模式

# Pseudo-code for cache-aside
def get_user(user_id):
    # Try cache first
    cached = redis.get(f"cache:user:{user_id}")
    if cached:
        return json.loads(cached)

    # Cache miss - fetch from database
    user = database.get_user(user_id)

    # Store in cache with expiration
    redis.setex(f"cache:user:{user_id}", 3600, json.dumps(user))

    return user

Write-Through 模式

def update_user(user_id, data):
    # Update database
    database.update_user(user_id, data)

    # Update cache
    redis.setex(f"cache:user:{user_id}", 3600, json.dumps(data))

缓存失效

# Delete specific cache
DEL cache:user:1234

# Delete by pattern (use with caution in production)
# Use SCAN instead of KEYS for large datasets
SCAN 0 MATCH cache:user:* COUNT 100

# Tag-based invalidation using sets
SADD cache:tags:user:1234 "cache:user:1234:profile" "cache:user:1234:orders"
# Invalidate all related caches
SMEMBERS cache:tags:user:1234
# Then delete each key

过期与内存管理

TTL 最佳实践

• 始终为缓存键设置 TTL
• 使用随机抖动，避免惊群效应
• 会话数据可考虑使用滑动过期

# Set with expiration
SET cache:data:123 "value" EX 3600

# Set expiration on existing key
EXPIRE cache:data:123 3600

# Check TTL
TTL cache:data:123

# Persist key (remove expiration)
PERSIST cache:data:123

内存管理

# Check memory usage
INFO memory

# Get key memory usage
MEMORY USAGE cache:large:object

# Configure max memory policy
CONFIG SET maxmemory 2gb
CONFIG SET maxmemory-policy allkeys-lru

事务与原子性

MULTI/EXEC 事务

# Transaction block
MULTI
INCR stats:views
LPUSH recent:views "page:123"
EXEC

# Watch for optimistic locking
WATCH user:1234:balance
balance = GET user:1234:balance
MULTI
SET user:1234:balance (balance - 100)
EXEC

Lua 脚本

• 用于复杂的原子操作
• 脚本以原子方式执行

-- Rate limiting script
local key = KEYS[1]
local limit = tonumber(ARGV[1])
local window = tonumber(ARGV[2])

local current = tonumber(redis.call('GET', key) or '0')

if current >= limit then
    return 0
end

redis.call('INCR', key)
if current == 0 then
    redis.call('EXPIRE', key, window)
end

return 1

# Execute Lua script
EVAL "return redis.call('GET', KEYS[1])" 1 mykey

发布/订阅与消息传递

# Publisher
PUBLISH channel:notifications '{"type":"alert","message":"New order"}'

# Subscriber
SUBSCRIBE channel:notifications

# Pattern subscription
PSUBSCRIBE channel:*

高可用性

复制

• 使用副本扩展读取能力
• 在主节点上配置恰当的持久化

# On replica
REPLICAOF master_host 6379

# Check replication status
INFO replication

Redis Sentinel

• 用于自动故障转移
• 至少部署 3 个 Sentinel 实例

Redis Cluster

• 用于水平扩展
• 数据会自动分片到各节点
• 为相关键使用哈希标签

# Hash tags ensure keys go to same slot
SET {user:1234}:profile "data"
SET {user:1234}:settings "data"

持久化

RDB 快照

# Manual snapshot
BGSAVE

# Configure automatic snapshots
CONFIG SET save "900 1 300 10 60 10000"

AOF（仅追加文件）

# Enable AOF
CONFIG SET appendonly yes
CONFIG SET appendfsync everysec

# Rewrite AOF
BGREWRITEAOF

安全

• 要求进行身份验证
• 使用 TLS 连接
• 绑定到指定接口
• 禁用危险命令

# Set password
CONFIG SET requirepass "your_strong_password"

# Authenticate
AUTH your_strong_password

# Rename dangerous commands (in redis.conf)
rename-command FLUSHALL ""
rename-command FLUSHDB ""
rename-command KEYS ""

监控

# Server info
INFO

# Memory stats
INFO memory

# Client connections
CLIENT LIST

# Slow log
SLOWLOG GET 10

# Monitor commands (debug only)
MONITOR

# Key count per database
INFO keyspace

连接管理

• 使用连接池
• 设置恰当的超时时间
• 妥善处理重新连接

# Python example with connection pool
import redis

pool = redis.ConnectionPool(
    host='localhost',
    port=6379,
    max_connections=50,
    socket_timeout=5,
    socket_connect_timeout=5
)

redis_client = redis.Redis(connection_pool=pool)

性能提示

• 使用流水线进行批量操作
• 避免使用大型键（值超过 100 KB）
• 在生产环境中使用 SCAN，而不是 KEYS
• 监控并优化内存使用情况
• 复杂的 JSON 操作可考虑使用 RedisJSON

# Pipeline example (pseudo-code)
pipe = redis.pipeline()
pipe.get("key1")
pipe.get("key2")
pipe.set("key3", "value")
results = pipe.execute()