- Upgrades `IPv6Dualstack` to `Beta` and turns it on by default. New clusters or existing clusters are not be affected until an actor starts adding secondary Pods and service CIDRS CLI flags as described here: [IPv4/IPv6 Dual-stack](https://github.com/kubernetes/enhancements/tree/master/keps/sig-network/563-dual-stack) ([#98969](htt...

`system:aggregate-to-edit` role no longer includes write access to the Endpoints API. For new Kubernetes 1.22 clusters, the `edit` and `admin` roles will no longer include that access in newly created Kubernetes 1.22 clusters. This will have no affect on existing clusters upgrading to Kubernetes 1.22. To retain write access to Endpoints in the aggregated `edit` and `admin` roles for newly created 1.22 clusters, refer to https://github.com/kubernetes/website/pull/29025. ([#103704](https://github.com/kubernet...

* Retry 'connection refused' errors when setting up clusters on GCE. ([#57394](https://github.com/kubernetes/kubernetes/pull/57394), [@mborsz](https://github.com/mborsz))
* Retry 'connection refused' errors when setting up clusters on GCE. ([#57394](https://github.com/kubernetes/kubernetes/pull/57394), [@mborsz](https://github.com/mborsz))

that they span multiple clusters. Creating a Federated HPA targeting
multiple clusters ensures that cluster-level autoscalers are
consistently deployed across those clusters, and dynamically managed
to ensure that autoscaling can occur optimially in all clusters,
within a set of global constraints on the total number of replicas
permitted across all clusters.  If replicas are not

- Fix computing backoff delay when using Job pod failure policy, by including in the backoff delay calculation pod failures ignored from the backoffLimit counter.
  
  Also, compute the backoff delay more accurately for deleted pods. ([#119466](https://github.com/kubernetes/kubernetes/pull/119466), [@mimowo](https://github.com/mimowo)) [SIG Apps]

  - *apiserver_cache_list_total*: Counter of LIST requests served from watch cache, broken down by resource_prefix and index_name
  - *apiserver_cache_list_fetched_objects_total*: Counter of objects read from watch cache in the course of serving a LIST request, broken down by resource_prefix and index_name

- Kube-apiserver metrics will now include request counts, latencies, and response sizes for /healthz, /livez, and /readyz requests. ([#83598](https://github.com/kubernetes/kubernetes/pull/83598), [@jktomer](https://github.com/jktomer)) [SIG API Machinery]

- On compatible systems, a mounter's Unmount implementation is changed to not return an error when the specified target can be detected as not a mount point. On Linux, the behavior of detecting a mount point depends on `umount` command is validated when the mounter is created. Additionally, mount point checks will be skipped in CleanupMountPoint/CleanupMountWithForce if the mounter's Unmount having the changed behavior of not returning error when target...

- Multiple Kubernetes clusters can now share the same resource group
    - When upgrading from previous releases, issues will arise with public IPs if multiple clusters share the same resource group. To solve these problems, make the following changes to the cluster:

* **[beta]** `kubefed` has graduated to beta: supports hosting federation on on-prem clusters, automatically configures `kube-dns` in joining clusters and allows passing arguments to federation components.

### Internal Storage Layer