Intermittent Redis Timeout in online Go Service with go-redis

[xindemeng]

Description

I’m experiencing sporadic Redis timeout issues in my Go service. These timeouts occur multiple times a day, each lasting approximately 2 seconds.

Environment

• Redis client library: github.com/redis/go-redis/v9@v9.6.0
• Redis operation frequency: High (many operations per second)

Observations

1. During the timeout period:
   • Stats.Misses and Stats.StaleConns show a sharp spike.
   • Stats.IdleConns still has a significant number of connections available.
2. Capturing network packets reveals that during the timeout, the client frequently closes connections and opens new ones. Notably, the client actively initiates connection closures.

Client Configuration

Here’s how the Redis client is initialized:

option := &redisv9.Options{  
    Addr:             "",  
    DialTimeout:      time.Second * 3,  
    ReadTimeout:      time.Millisecond * 300,  
    Password:         "",   
    DB:               0,  
    DisableIndentity: true,  
    PoolSize:         200,  
}  
client := redisv9.NewClient(option)  

Questions and Thoughts

• The connection pool (PoolSize: 200) seems sufficient given the operation frequency, and connections should not expire simultaneously.
• Could the issue be related to the connCheck method used for connection health checks? It seems plausible, but I’m unsure how to verify or debug this.

I’m open to any ideas or suggestions to help diagnose or resolve this issue. Your insights would be greatly appreciated!