Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Prombench: tweak rw-sink resource usage #794

Merged
merged 1 commit into from
Nov 27, 2024
Merged

Prombench: tweak rw-sink resource usage #794

merged 1 commit into from
Nov 27, 2024

Conversation

bboreham
Copy link
Member

Since the sink program handles a lot of messages but doesn't do anything with them, it does a lot of garbage-collection.
Tell the Go runtime we would like it to use more memory, which reduces the CPU by about half.

Adjust requests to match.

image

Copy link
Member

@bwplotka bwplotka left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Wow, amazing thanks!

How much traffic do you send it to it, so it consumes 4 CPU? Just Prometheus we "prombench"? I think we might want to limit concurrency to give more predictable resource usage, sink does not need to that "fast" in responding perhaps?

@bwplotka bwplotka merged commit 2b19c16 into master Nov 27, 2024
4 checks passed
@bwplotka bwplotka deleted the tweak-rwsink branch November 27, 2024 13:36
@bboreham
Copy link
Member Author

This is 1M series scraped every 5s, on two Prometheus. So 400K samples/sec?

I'm not very sure what you mean about limiting concurrency. We use 8-core nodes, so it won't be more than that.

@bwplotka
Copy link
Member

I mean we can add limit of the concurrency on sink, so it receives slower, but with stable CPU use. But you are right that a lot of it was through GC.

I will also play with memory regions (new proposal) to reduce GC. Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants