Skip to content

Enhance node metrics caching #1736

Description

@ImNotAFrog

What would you like to be added:
Support multiple requests in one resolution period.
Why is this needed:
I found that metrics-server does not truly cache node metrics. For example, with high-loaded cluster, part of cluster's kubelets' requests went timeout during each metrics collection (varying nodes each time) — leaving metrics-server almost unable to provide a complete set of node metrics, regardless of how the resolution is adjusted.

PS: If this proposal is adopted, I’d greatly appreciate the opportunity to participate in the modifications.

/kind feature

Metadata

Metadata

Assignees

Labels

kind/featureCategorizes issue or PR as related to a new feature.triage/acceptedIndicates an issue or PR is ready to be actively worked on.

Type

No type

Fields

No fields configured for issues without a type.

Projects

Status
Needs Triage

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions