Potential memory leaks on 9.12.0

[onizmx]

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

Expected Behavior

I would not expect this steep increase of memory consumption.

Current Behavior

Soon after it got deployed, ECS memory consumption got increased from 2% to 10%.
A Lamba function with 128mb of memory also failed several invocations.
Might not have enough data points since service deployment has been reverted, but looks like it's also gradually increasing.

Possible Solution

Don't have one.

Steps to Reproduce

Deployed our service which handles high throughput of event streams.

Context (Environment)

go 1.24.5
ECS: ARM64, debian:12-slim
Lambda: x86_64, provided.al2023

Event stream Lambda handlers got behind, backlogs built, ECS memory spike or leak didn't cause a problem since it was reverted.

Detailed Description

Possible Implementation

[ndyakov]

Hello @onizmx, thank you for reporting this! Can you share which was the previous version you were using?

[onizmx]

@ndyakov Hi!
We were using 9.11.0 which I assume just a version before?

[ndyakov]

@onizmx Correct, I am just looking at the diff at the moment and trying to find the reason. One thing that looks promising is the new buffer size configurations: https://github.com/redis/go-redis?tab=readme-ov-file#buffer-size-configuration
Can you pleaase try by setting the buffer to a smaller size, previously the default as the default of go - 4096 https://github.com/golang/go/blob/master/src/bufio/bufio.go#L19 and now is 524288, which will make sense why you see the initial increase.

Feel free to drop it back to 4096 with the configuration and check if you observe any increase in the usage that look like potential memory leak.

[onizmx]

@ndyakov Thank you! I will be away for the next few days so will try it early next week and give you an update!
Just to make sure, these will be the configuration I will try:

	client := redis.NewClusterClient(&redis.ClusterOptions{
		Addrs:            []string{address},
		DisableIndentity: true,
		ReadBufferSize:   4096, <-- new
		WriteBufferSize:  4096, <-- new
	})
	readOnlyClient := redis.NewClusterClient(&redis.ClusterOptions{
		Addrs:            []string{address},
		ReadOnly:         true,
		RouteRandomly:    true,
		RouteByLatency:   true,
		DisableIndentity: true,
		ReadBufferSize:   4096, <-- new
		WriteBufferSize:  4096, <-- new
	})

[ndyakov]

Yes @onizmx, this should use the same size as in 9.11 , keep in mind bigger buffers can be beneficial for speed with bigger commands. You can play around with the size while testing. Please let me know when you have more data and we can decide how to proceed.