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

HTTP Connection Manager: Add the ability to have per worker stats #33671

Open
KBaichoo opened this issue Apr 18, 2024 · 2 comments
Open

HTTP Connection Manager: Add the ability to have per worker stats #33671

KBaichoo opened this issue Apr 18, 2024 · 2 comments
Assignees
Labels
area/stats enhancement Feature requests. Not bugs or questions. no stalebot Disables stalebot from closing an issue

Comments

@KBaichoo
Copy link
Contributor

HTTP Connection Manager: Add the ability to have per worker stats

Description:

Add stats that will allow us to see imbalance among workers. This is particularly useful for mesh scenarios wherein you might use an exact balance to balance the connection but can still end up with imbalance at the request level.

[optional Relevant Links:]

Any extra documentation required to understand the issue.

@KBaichoo KBaichoo added enhancement Feature requests. Not bugs or questions. triage Issue requires triage labels Apr 18, 2024
@KBaichoo KBaichoo self-assigned this Apr 18, 2024
@adisuissa adisuissa added no stalebot Disables stalebot from closing an issue area/stats and removed triage Issue requires triage labels Apr 19, 2024
@adisuissa
Copy link
Contributor

I'll let you cc relevant people.

Just thinking out loud here:

This is particularly useful for mesh scenarios wherein you might use an exact balance to balance the connection but can still end up with imbalance at the request level.

In H2 I assume that there may be imbalance at the request level. So adding stats is useful to keep track of things, but not sure if they can be used to change balancing.

@KBaichoo
Copy link
Contributor Author

KBaichoo commented Apr 22, 2024

Worker imbalance is a bit hard to see, the H2 example that you provide is another case. The reason why this is useful for service mesh is operators might be able to influence the callers compared to say edge use cases.

Anyhow, allowing us to see the imbalance problem can help with figuring out the solution that makes sense in a given case:

  • Maybe the solution in some cases is to change how callers call
  • Maybe the solution is to limit the max requests per stream lifetime
  • Maybe the solution is an endpoint that does GOAWAY to try to rebalance.

cc @alyssawilk @jmarantz

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
area/stats enhancement Feature requests. Not bugs or questions. no stalebot Disables stalebot from closing an issue
Projects
None yet
Development

No branches or pull requests

2 participants