Multi vCenter support is deprecated as of v1.21. If you have a Kubernetes cluster spanning across multiple vCenter servers, please consider moving all k8s nodes to a single vCenter Server. vSphere CSI Driver does not support Kubernetes deployment spanning across multiple vCenter servers.
  

- 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

* Timeout docker queries to prevent node going NotReady indefinitely. ([#56967](https://github.com/kubernetes/kubernetes/pull/56967), [@jsravn](https://github.com/jsravn))
* Fix bug allowing garbage collector to enter a broken state that could only be fixed by restarting the controller-manager. ([#61201](https://github.com/kubernetes/kubernetes/pull/61201), [@jennybuckley](https://github.com/jennybuckley))

  So, if you want to just keep the existing behavior, you can register `ClusterEventWithHint` with no `QueueingHintFn` in it. 
  But, registering appropriate `QueueingHintFn` is, of course, better from a scheduling performance perspective. ([#118551](https://github.com/kubernetes/kubernetes/pull/118551), [@sanposhiho](https://github.com/sanposhiho)) [SIG Node, Scheduling, Storage and Testing]

seccomp defaults, the kubelet uses the `RuntimeDefault` seccomp profile by default, rather than than `Unconfined`. This allows enhancing the default cluster wide workload security of the Kubernetes deployment. Security administrators will now sleep better knowing there is some security by default for the workloads.

To learn more about the feature, please refer to the official [seccomp tutorial](https://kubernetes.io/docs/tutorials/clusters/seccomp/#enable-the-use-of-runtimedefault-as-...

leads to Node goes `Not-ready` state when credentials for vCenter stored in a secret and Zones feature is in use. Zone labels setup moved to KCM component, kubelet skips this step during startup in such case. If credentials stored in cloud-provider config file as plaintext current behaviour does not change and no action required. For proper functioning `kube-system:vsphere-legacy-cloud-provider` should be allowed to update node object if vCenter credentials stored in secret and Zone feature used. ([...

### Other (Cleanup or Flake)

- Service session affinity timeout tests are no longer required for Kubernetes network plugin conformance due to variations in existing implementations. New conformance tests will be developed to better express conformance in future releases. ([#112806](https://github.com/kubernetes/kubernetes/pull/112806), [@dcbw](https://github.com/dcbw)) [SIG Architecture, Network and Testing]

## Dependencies

### Added

- All `klog` flags except for `-v` and `-vmodule` are deprecated. Support for `-vmodule` is only guaranteed for the text log format. ([#105042](https://github.com/kubernetes/kubernetes/pull/105042), [@pohly](https://github.com/pohly))

- kubeadm: Added better handling of errors during unmount when calling `kubeadm reset`. When failing to unmount directories under `/var/run/kubelet`, kubeadm will now throw an error instead of showing a warning and continuing to clean up said directory. In such situations, it is better for you to inspect the problem and resolve it manually. Then, you can call `kubeadm reset` again to complete the cleanup....

  you want to set the equivalent of any of the `--iptables-xxx` options.)
  
  As this code is still very new, it has not been heavily optimized yet;
  while it is expected to _eventually_ have better performance than the
  iptables backend, very little performance testing has been done so far. ([#121046](https://github.com/kubernetes/kubernetes/pull/121046), [@danwinship](https://github.com/danwinship))