connection pool timeout - not reconnect

[parsibox]

hi
i use go 1.23 on linux 64 bit
today my redis has a problem and i restarted that but my golang code have error and not reconnected

this is my code

	RedisClient = redis.NewClient(&redis.Options{
		Addr:     ip + ":" + port, //"localhost:6379",
		Password: password,
		DB:       db,
		PoolSize: poolSize,
		ConnMaxLifetime: time.Minute * 10,
		DialTimeout:     time.Second * 10,
		ReadTimeout:     time.Second * 30,
		WriteTimeout:    time.Second * 10,
		MaxRetries:      3000,
		MinRetryBackoff: time.Second * 2,
		MaxRetryBackoff: time.Second * 10,
		OnConnect: func(ctx context.Context, cn *redis.Conn) error {
			logger.Info("ConRedis connected")
			return nil
		},
	})
	if err := RedisClient.Ping(context.Background()).Err(); err != nil {
		// Handle connection error
		logger.Error("ConRedis error")
	}

i get error : redis: connection pool timeout

why pool connection not renew?or not reconnect again?after i restarted my code it work !!!

[ndyakov]

Hello @parsibox , what is the poolSize that you are using?

[parsibox]

30
my redis only down for 15 min and after that i do not have any good connection on pool
all connection say "connection pool timeout" i think they should reconnect and work again

[monkey92t]

redis-server version?
go-redis version?

[y-kawawa]

I also encountered an issue where connection pool timeouts were not subject to retries.

• client
  • go 1.23.7
  • go-redis v9.7.1
• server
  • Valkey serverless engine version: 8

opts, _ := redis.ParseURL(url)
opts.DialTimeout = 1 * time.Second
opts.ReadTimeout = 1 * time.Second
opts.WriteTimeout = 1 * time.Second
opts.MaxRetries = 2

I suspect that the reason connection pool timeouts are not being retried is as follows:
ErrPoolTimeout is excluded in shouldRetry.

[monkey92t]

I also encountered an issue where connection pool timeouts were not subject to retries.

• client

  • go 1.23.7
  • go-redis v9.7.1

• server

  • Valkey serverless engine version: 8

opts, _ := redis.ParseURL(url)
opts.DialTimeout = 1 * time.Second
opts.ReadTimeout = 1 * time.Second
opts.WriteTimeout = 1 * time.Second
opts.MaxRetries = 2
I suspect that the reason connection pool timeouts are not being retried is as follows: ErrPoolTimeout is excluded in shouldRetry.

Indeed, this error was overlooked, and it can be fixed.

[monkey92t]

I also encountered an issue where connection pool timeouts were not subject to retries.

• client

  • go 1.23.7
  • go-redis v9.7.1

• server

  • Valkey serverless engine version: 8

opts, _ := redis.ParseURL(url)
opts.DialTimeout = 1 * time.Second
opts.ReadTimeout = 1 * time.Second
opts.WriteTimeout = 1 * time.Second
opts.MaxRetries = 2
I suspect that the reason connection pool timeouts are not being retried is as follows: ErrPoolTimeout is excluded in shouldRetry.

#3298

[parsibox]

redis-server version? go-redis version?

Redis Version: 7.0.11
OS: Linux5.4.0-125-genericx86_64
github.com/redis/go-redis/v9 v9.7.1

[monkey92t]

redis-server version? go-redis version?

Redis Version: 7.0.11 OS: Linux5.4.0-125-genericx86_64 github.com/redis/go-redis/v9 v9.7.1

It is observed that you are using the latest version of go-redis and encountered the error "redis: connection pool timeout." This error corresponds to ErrPoolTimeout in go-redis (ErrPoolTimeout timed out waiting to get a connection from the connection pool), indicating that go-redis timed out while trying to obtain a connection from the connection pool. This suggests that there are no available connections in the pool, and the PoolSize limit has been reached. The timeout duration depends on opt.PoolTimeout (if not set, it defaults to ReadTimeout + 1s).

Your PoolSize is set to 30. From the scenario above, it can be inferred that your connection pool is exhausted. The connections might be waiting for read/write operations or could have been manually manipulated (e.g., via the Conn function). Could you provide an example that reproduces this issue?

[parsibox]

redis-server version? go-redis version?

Redis Version: 7.0.11 OS: Linux5.4.0-125-genericx86_64 github.com/redis/go-redis/v9 v9.7.1

It is observed that you are using the latest version of go-redis and encountered the error "redis: connection pool timeout." This error corresponds to ErrPoolTimeout in go-redis (ErrPoolTimeout timed out waiting to get a connection from the connection pool), indicating that go-redis timed out while trying to obtain a connection from the connection pool. This suggests that there are no available connections in the pool, and the PoolSize limit has been reached. The timeout duration depends on opt.PoolTimeout (if not set, it defaults to ReadTimeout + 1s).  Your PoolSize is set to 30. From the scenario above, it can be inferred that your connection pool is exhausted. The connections might be waiting for read/write operations or could have been manually manipulated (e.g., via the Conn function). Could you provide an example that reproduces this issue?

i only first run a script in redis and that was heavy then i restart redis after that my redis pool connection get connection pool timeout
i think you should check first is there any live connection in connection pool or not if there is not then remove old connection from pool and create new connection in pool

[parsibox]

or maybe there is a bug in remove died connection from pool

[monkey92t]

We cannot delete connections in the connection pool; they might be in use (or even in use for a long time), which is one of the connection pool's functions. Based on what you're describing, I believe heavy requests are causing the connection pool to be very busy. Am I understanding this correctly?

[joe-ton]

Hey @parsibox, @monkey92t — I'd like to look into contributing here. I'm particularly interested in improving the robustness of the connection pool logic, specifically around handling timeouts, reconnect attempts, and stale connection management.

I've reviewed the discussion above and the details provided, and I'm ready to dig deeper into this. If there are specific areas of the codebase you think would benefit most from attention, please let me know. Otherwise, I'll start by investigating the handling of ErrPoolTimeout in the retry and connection pool logic.

Looking forward to helping here!

[ndyakov]

@parsibox could you be able to test with the latest 9.8.0-beta.1 ?