## Changes by Kind

### Deprecation

  the endpoints for it by hand (rather than letting kube-controller-manager
  generate `Endpoints` and `EndpointSlices` for it automatically based on the
  Service definition), then you should write out an EndpointSlice object rather
  than (or in addition to) an `Endpoints` object. ([#129837](https://github.com/kubernetes/kubernetes/pull/129837), [@danwinship](https://github.com/danwinship)) [SIG API Machinery, Network and Testing]

- Improved performance of `Endpoint` and `EndpointSlice` controllers when there are a large number of services in a single namespace by making pod-to-service lookup asynchronous. ([#134739](https://github.com/kubernetes/kubernetes/pull/134739), [@shyamjvs](https://github.com/shyamjvs)) [SIG Apps and...

- Promoted several endpointslice metrics from Alpha to Beta stability. ([#136368](https://github.com/kubernetes/kubernetes/pull/136368), [@bhope](https://github.com/bhope)) [SIG Instrumentation and Network]

- Fixed: during the kube-controller-manager restart, when the corresponding Endpoints resource was manually deleted and recreated, causing the endpointslice to fail to be created normally. ([#125359](https://github.com/kubernetes/kubernetes/pull/125359), [@yangjunmyfm192085](https://github.com/yangjunmyfm192085)) [SIG Apps and Network]

- The following roles have had `Watch` added to them (prefixed with `system:controller:`):
  
  - `cronjob-controller`
  - `endpoint-controller`
  - `endpointslice-controller`
  - `endpointslicemirroring-controller`
  - `horizontal-pod-autoscaler`
  - `node-controller`
  - `pod-garbage-collector`