- Fixed bug that caused a resource to include patch directives when using strategic merge patch against a non-existent field. ([#117568](https://github.com/kubernetes/kubernetes/pull/117568), [@alexzielenski](https://github.com/alexzielenski))

empty cgroupDriver value in the KubeletConfiguration is now always set to "systemd" unless the user is explicit about it. This requires existing machine setups to configure the container runtime to use the "systemd" driver. Documentation on this topic can be found here: https://kubernetes.io/docs/setup/production-environment/container-runtimes/. When upgrading existing clusters / nodes using "kubeadm upgrade" the old cgroupDriver value is preserved, but in 1.22 this change will also apply to "upgrade"....

- Added new option to the `InterPodAffinity` scheduler plugin to ignore existing
  pods` preferred inter-pod affinities if the incoming pod has no preferred inter-pod
  affinities. This option can be used as an optimization for higher scheduling throughput
  (at the cost of an occasional pod being scheduled non-optimally/violating existing
  pods preferred inter-pod affinities). To enable this scheduler option, set the

* Add --no-negcache flag to kube-dns to prevent caching of NXDOMAIN responses. ([#66036](https://github.com/kubernetes/kubernetes/pull/66036), [@prameshj](https://github.com/prameshj))
* fix data loss issue if using existing azure disk with partitions in disk mount  ([#63270](https://github.com/kubernetes/kubernetes/pull/63270), [@andyzhangx](https://github.com/andyzhangx))

### Beta APIs Off by Default

[New beta APIs will not be enabled in clusters by default](https://github.com/kubernetes/enhancements/issues/3136).
Existing beta APIs and new versions of existing beta APIs, will continue to be enabled by default.

### Signing Release Artifacts

CSI Migration enables the replacement of existing in-tree storage plugins such as `kubernetes.io/gce-pd` or `kubernetes.io/aws-ebs` with a corresponding CSI driver.
If CSI Migration is working properly, Kubernetes end users shouldn’t notice a difference.
After migration, Kubernetes users may continue to rely on all the functionality of in-tree storage plugins using the existing interface.

- Applying fix for not deleting existing public IP when a service is deleted in Azure. ([#100694](https://github.com/kubernetes/kubernetes/pull/100694), [@nilo19](https://github.com/nilo19))
- Applying fix for not tagging static public IP. ([#101752](https://github.com/kubernetes/kubernetes/pull/101752), [@nilo19](https://github.com/nilo19))

EndpointSlices with Pod referencing Nodes that don't exist couldn't be created or updated. The behavior on the EndpointSlice controller has been modified to update the EndpointSlice without the Pods that reference non-existing Nodes and keep retrying until all Pods reference existing Nodes. However, if `service.Spec.PublishNotReadyAddresses` is set, all the Pods are published without retrying. Fixed EndpointSlices metrics to reflect correctly the number of desired EndpointSlices when no endpoints...

   ([#123372](https://github.com/kubernetes/kubernetes/pull/123372), [@carl...

### API Change