- The label applied to control-plane nodes "node-role.kubernetes.io/master" is now deprecated and will be removed in a future release after a GA deprecation period.
  - Introduce a new label "node-role.kubernetes.io/control-plane" that will be applied in parallel to "node-role.kubernetes.io/master" until the removal of the "node-role.kubernetes.io/master" label.

  - `GCERegionalPersistentDisk`

## Metrics Changes

### Added metrics

- The metric `kube_proxy_sync_proxy_rules_last_timestamp_seconds` is now available, indicating the last time that kube-proxy successfully applied proxying rules. ([#74027](https://github.com/kubernetes/kubernetes/pull/74027), [@squeed](https://github.com/squeed))

- The alpha feature `ConfigurableFSGroupPolicy` enables v1 Pods to specify a spec.securityContext.fsGroupChangePolicy policy to control how file permissions are applied to volumes mounted into the pod. ([#88488](https://github.com/kubernetes/kubernetes/pull/88488), [@gnufied](https://github.com/gnufied)) [SIG  Storage]

- Introduced dynamic volume provisioning shim for CSI migration ([#73653](https://github.com/kubernetes/kubernetes/pull/73653), [@ddebroy](https://github.com/ddebroy))
- Applied resources from a directory containing kustomization.yaml ([#74140](https://github.com/kubernetes/kubernetes/pull/74140), [@Liujingfang1](https://github.com/Liujingfang1))

#### Defaults

 - The default spec.updateStrategy for the StatefulSet and DaemonSet kinds is
 RollingUpdate for the apps/v1beta2 group version. You can explicitly set
 the OnDelete strategy, and no strategy auto-conversion is applied to
 replace default values.

 - As mentioned in [Behavioral Changes](#behavioral-changes), selector
 defaults are disabled.

 - The default spec.revisionHistoryLimit for all applicable kinds in the

- Kubeadm: Remove the deprecated hyperkube image support in `v1beta3`. This implies removal of `ClusterConfiguration.UseHyperKubeImage.` ([#101537](https://github.com/kubernetes/kubernetes/pull/101537), [@neolit123](https://github.com/neolit123))

- Kubeadm: deprecate the feature `--experimental-kustomize` in favor of `--experimental-patches`. The supported patch formats are the same as "kubectl patch". They are read as files from a directory and can be applied to kubeadm components during init/join/upgrade. Only patching of static Pods is supported for the time being. ([#92017](https://github.com/kubernetes/kubernetes/pull/92017), [@neolit123](https://github.com/neolit123))

- Simplied validation error message for invalid fields by removing redundant field name. ([#132513](https://github.com/kubernetes/kubernetes/pull/132513), [@xiaoweim](https://github.com/xiaoweim)) [SIG API Machinery, Apps, Auth, Node and Scheduling]

- Enhanced cloud provider integrations to support optional, per-Node custom labels that can be supplied and applied to Nodes by the node controller. These extra labels will only be applied where the cloud provider integration supports this feature.
   ([#123223](https://github.com/kubernetes/kubernetes/pull/123223), [@mmerkes](https://github.com/mmerkes))

    *   Losses are scaled in canned estimator v2 and not in the optimizers
        anymore. If you are using Estimator + distribution strategy + optimikzer
        v1 then the behavior does not change. This implies that if you are using
        custom estimator with optimizer v2, you have to scale losses. We have
        new utilities to help scale losses `tf.nn.compute_average_loss`,
        `tf.nn.scale_regularization_loss`.