Skip to content

GEP: Standardized Telemetry API #4768

@gkhom

Description

@gkhom

What would you like to be added:

I would like to propose the addition of a standardized Telemetry API (initially drafted as TelemetryPolicy) to the Gateway API project. This API aims to provide a unified, provider-agnostic way to configure observability signals (metrics, access logs, traces) for both North/South (Gateway) and East/West (Mesh) traffic. My request is to initiate a Provisional GEP.

Why this is needed:

The current Kubernetes landscape suffers from "observability lock-in" because the configuration for telemetry is deferred to vendor-specific APIs. This fragmentation makes it difficult for platform owners to maintain consistent observability across different infrastructure providers and implementations.

Furthermore, emerging workloads like AI Agents elevate the criticality of consistent observability significantly due to their non-deterministic nature in which runtime behavior is not dictated by the workload but rather by an external LLM.

A standardized Telemetry API within the Gateway API ecosystem will:

  1. Decouple the intent of observability from the underlying vendor-specific implementation.
  2. Provide uniform configurability for both Gateway and Mesh use cases.
  3. Account for the fact that the persona responsible for the Gateway/Mesh infrastructure can be different from the persona responsible for dictating the structure and behavior of telemetry signals.
  4. Empower platform and observability teams to enforce uniform telemetry standards across large-scale heterogeneous environments.
  5. Ensure the API is suitable for the broader ecosystem beyond specialized networking.

This proposal initially started in the Kube-Agentic-Networking subproject. The community deemed the proposal sufficiently generic to be moved to the Gateway API instead. The discussion and feedback can be found here: kubernetes-sigs/kube-agentic-networking#69

Metadata

Metadata

Assignees

Labels

kind/featureCategorizes issue or PR as related to a new feature.kind/gepPRs related to Gateway Enhancement Proposal(GEP)triage/acceptedIndicates an issue or PR is ready to be actively worked on.

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions