TxPipeline.Exec timeout when there is command failed to be enqueued

[howz97]

Issue tracker is used for reporting bugs and discussing new features. Please use
stackoverflow for supporting issues.

TxPipeline.Exec timeout when there is command failed to be enqueued. I used go-redis to connect a distributed redis cluster, The redis cluster require that all keys of a transaction must located in the same slot, otherwise the subsequent command can not be added into the transaction queue.

This is my program:

package main

import (
	"context"
	"fmt"

	"github.com/redis/go-redis/v9"
)

func main() {
	ctx := context.Background()

	client := redis.NewClient(&redis.Options{
		Addr: "localhost:62908",
	})
	defer client.Close()

	fmt.Println("Use TxPipeline ...")
	pipe := client.TxPipeline()
	if err := pipe.Set(ctx, "a", 1, 0).Err(); err != nil {
		fmt.Printf("TxPipeline push error: %v\n", err)
	}
	if err := pipe.Set(ctx, "b", 1, 0).Err(); err != nil {
		fmt.Printf("TxPipeline push error: %v\n", err)
	}
	cmders, err := pipe.Exec(ctx)
	if err != nil {
		fmt.Printf("TxPipeline Exec Error: %v\n", err)
		for _, cmder := range cmders {
			if cmder.Err() != nil {
				fmt.Printf("  cmd=%q err=%v\n", cmder.String(), cmder.Err())
			}
		}
	} else {
		fmt.Println("TxPipeline succeed")
	}

	fmt.Println("\nUse MULTI/EXEC ...")
	if err := client.Do(ctx, "MULTI").Err(); err != nil {
		fmt.Printf("MULTI error: %v\n", err)
		return
	}
	if err := client.Set(ctx, "a", 1, 0).Err(); err != nil {
		fmt.Printf("MULTI push error: %v\n", err)
	}
	if err := client.Set(ctx, "b", 1, 0).Err(); err != nil {
		fmt.Printf("MULTI push error: %v\n", err)
	}
	result, err := client.Do(ctx, "EXEC").Result()
	if err != nil {
		fmt.Printf("MULTI EXEC error: %v\n", err)
	} else {
		fmt.Printf("MULTI EXEC result: %v\n", result)
	}
}

The output is:

Use TxPipeline ...
TxPipeline Exec Error: read tcp 127.0.0.1:20204->127.0.0.1:62908: i/o timeout
  cmd="set a 1: read tcp 127.0.0.1:20204->127.0.0.1:62908: i/o timeout" err=read tcp 127.0.0.1:20204->127.0.0.1:62908: i/o timeout
  cmd="set b 1: read tcp 127.0.0.1:20204->127.0.0.1:62908: i/o timeout" err=read tcp 127.0.0.1:20204->127.0.0.1:62908: i/o timeout

Use MULTI/EXEC ...
MULTI push error: ERR in transaction context, keys must in same slot
MULTI EXEC result: [OK]

Expected Behavior

TxPipeline.Exec should report the same error as MULTI/EXEC: 'ERR in transaction context, keys must be in the same slot'.

Current Behavior

TxPipeline.Exec timeout

Possible Solution

Steps to Reproduce

Context (Environment)

Detailed Description

Possible Implementation

[ofekshenawa]

Hey @howz97,
Could you share a bit more information to help confirm the exact behavior?

• Which go-redis version are you using?
• What Redis/cluster/proxy implementation is running on localhost:62908?
• Does the same example still timeout if you use redis.NewClusterClient instead of redis.NewClient?
• If possible, can you provide a minimal Docker/cluster setup that reproduces this?

[howz97]

Thanks for your reply @ofekshenawa

The version of go-redis i am using is v9.19.0
localhost:62908 is the proxy address of baidu distributed redis cluster, and can't be connected with NewClusterClient.

There is a key different between the proxy of baidu redis and official redis is that the former did not report EXECABORT Transaction discarded because of previous errors when there is any command failed to be enqueued.