redis-safety

Redis Security Guidelines

Prevent common Redis performance and stability issues, applicable to all languages.

---

Prohibited Operations

Prohibited	Alternative	Reason
KEYS * / KEYS pattern	SCAN Cursor iteration	KEYS is an O(N) blocking operation and will cause Redis to freeze in production environments
FLUSHDB / FLUSHALL	SCAN + DEL by key prefix	Full deletion carries extremely high risk
SET without TTL	All keys must have a TTL set	Avoid memory leaks

Must Comply

1. Use SCAN instead of KEYS

// Go (go-redis) ❌
keys, _ := rdb.Keys(ctx, "user:*").Result()

// Go (go-redis) ✅
var cursor uint64
for {
    keys, cursor, _ = rdb.Scan(ctx, cursor, "user:*", 100).Result()
    // 处理 keys
    if cursor == 0 { break }
}

// Java (Jedis) ❌
Set<String> keys = jedis.keys("user:*");

// Java (Jedis) ✅
ScanParams params = new ScanParams().match("user:*").count(100);
String cursor = "0";
do {
    ScanResult<String> result = jedis.scan(cursor, params);
    // 处理 result.getResult()
    cursor = result.getCursor();
} while (!cursor.equals("0"));

# Python (redis-py) ❌
keys = r.keys("user:*")

# Python (redis-py) ✅
for key in r.scan_iter(match="user:*", count=100):
    # 处理 key

2. Large key control

• The value of a single key must not exceed 10KB
• Collection types (List/Set/Hash/ZSet) must not exceed 5000 elements
• Split into multiple keys when exceeded

3. Pipeline batch operations

Multiple Redis calls should use Pipeline to reduce network round trips:

// ❌ 循环单次调用
for _, id := range ids {
    rdb.Get(ctx, "user:"+id)
}

// ✅ Pipeline 批量
pipe := rdb.Pipeline()
for _, id := range ids {
    pipe.Get(ctx, "user:"+id)
}
pipe.Exec(ctx)

4. Set TTL for all keys

// ❌
rdb.Set(ctx, "token:123", value, 0)

// ✅
rdb.Set(ctx, "token:123", value, 24*time.Hour)