Deployment objects may cause the deployment controller to mistakenly match
other pods and delete them, due to the selector API change.
     * Client (kubectl) and server versions must match (both 1.1 or both 1.2) for any
Deployment-related operations.
     * Behavior change:
        * Deployment creates ReplicaSets instead of ReplicationControllers.

  an external cloud provider that has been updated to support the feature,
  you can pass comma-separated dual-stack `--node-ips` to `kubelet` and have
  the cloud provider take both IPs into account. ([#120275](https://github.com/kubernetes/kubernetes/pull/120275), [@danwinship](https://github.com/danwinship))

* Fix a problem of not respecting TerminationGracePeriodSeconds of the Pods created by DaemonSet controller. ([#51279](https://github.com/kubernetes/kubernetes/pull/51279), [@kow3ns](https://github.com/kow3ns))
* BUG FIX: Check both name and ports for azure health probes ([#56918](https://github.com/kubernetes/kubernetes/pull/56918), [@feiskyer](https://github.com/feiskyer))

- The apiserver, including both the kube-apiserver and apiservers built with the generic apiserver library, will now return 413 RequestEntityTooLarge error if a json patch contains more than 10,000 operations. ([#74000](https://github.com/kubernetes/kubernetes/pull/74000), [@cae...

  The possibility of borrowing means that the old metric `apiserver_flowcontrol_request_concurrency_limit` can no longer mean both the configured concurrency limit and the enforced concurrency limit. Henceforth it means the configured concurrency limit.' ([#113485](https://github.com/kubernetes/kubernetes/pull/113485), [@MikeSpreitzer](https://github.com/MikeSpreitzer))