Overview

Aggregates discovery into a small number of resources so that clients do not need make multiple requests to obtain complete discovery information.

Feature Availability

• Kubernetes:
  • Alpha: v1.26
  • Beta: v1.27
• OpenShift:
  • Alpha: v4.13
  • Beta: v4.14

Exploitation

The DiscoveryClient will default to using aggregated discovery once the feature is in beta. We simply need to update clients to the latest version of k8s.io/client-go to pick up the new behavior.

The DiscoveryClient can be configured to not use aggregation, giving us an option, if needed if aggregated discovery causes issues with an existing client.

Updates

06/20/23 - Vu

The Aggregated Discovery feature is enabled for Kubernetes 1.27 (as beta) by default and also enabled by default in OCP 4.14. This feature is still under feature gate even it is enabled by default. So it is possible to disable this feature via feature gate flag.

For aggregated APIs such as OpenShift-specific APIs that are controlled by openshift-apiserver, the aggregated discovery is enabled via apiserver library that we use. The important config setting for aggregated discovery is the `apis` endpoint which is enabled by this field in the apiserver config. It is important to remember not to overwritten or unset that field by mistakes.

References

• KEP-3352: Aggregated Discovery