### Breaking Changes


`tf.keras`:

addition to `--grace-period=0`.  This prevents users from accidentally force deleting pods without being aware of the consequences of force deletion.  Force deleting pods for resources like StatefulSets can result in multiple pods with the same name having running processes in the cluster, which may lead to data corruption or data inconsistency when using shared storage or common API endpoints. ([#35484](https://github.com/kubernetes/kubernetes/pull/35484), [@smarterclayton](https://github.com/smarterclayton))...

- Fixed a regression introduced in 1.22.4 where Azure load balancers were not kept up to date with the state of cluster nodes. In particular, nodes that are not in the ready state and are not newly created (i.e. not having the `node.cloudprovider.kubernetes.io/uninitialized` taint) now get removed from Azure load balancers. ([#109933](https://github.com/kubernetes/kubernetes/pull/109933), [@ricky-rav](https://github.com/ricky-rav)) [SIG Cloud Provider]

graduated the `UnversionedKubeletConfigMap` feature gate to Beta and enabled the feature by default. This implies that 1) for new clusters kubeadm will start using the `kube-system/kubelet-config` naming scheme for the kubelet ConfigMap and RBAC rules, instead of the legacy `kubelet-config-x.yy` naming. 2) during upgrade, kubeadm will only write the new scheme ConfigMap and RBAC objects. To disable the feature you can pass `UnversionedKubeletConfigMap: false` in the kubeadm config for new clusters....

- The `ClusterTrustBundle` API is moving to `v1beta1`.

- Fixed an issue where requests sent by the KMSv2 service would be rejected due to having an invalid authority header. ([#126930](https://github.com/kubernetes/kubernetes/pull/126930), [@Ruddickmg](https://github.com/Ruddickmg)) [SIG API Machinery and Auth]

  
  For developers of out-of-tree PostFilter plugins, note that the semantics of NodeToStatusMap are changing: A node with an absent value in the NodeToStatusMap should be interpreted as having an UnschedulableAndUnresolvable status. ([#125197](https://github.com/kubernetes/kubernetes/pull/125197), [@gabesaba](https://github.com/gabesaba))

	err = c.validateIDPSettings(ctx, sites)
	if err != nil {
		return madmin.ReplicateAddStatus{}, err
	}

	// For this `add` API, either all clusters must be empty or the local
	// cluster must be the only one having some buckets.
	if localHasBuckets && nonLocalPeerWithBuckets != "" {
		return madmin.ReplicateAddStatus{}, errSRInvalidRequest(errors.New("only one cluster may have data when configuring site replication"))
	}

        * `workqueue_work_duration_seconds`
        * `workqueue_unfinished_work_seconds`
        * `workqueue_longest_running_processor_seconds`
        * `workqueue_retries_total`
* Fix inability to use k8s with dockerd having default IPC mode set to private. ([#70826](https://github.com/kubernetes/kubernetes/pull/70826), [@kolyshkin](https://github.com/kolyshkin))

- Fixed a regression introduced in 1.23.0 where Azure load balancers were not kept up to date with the state of cluster nodes. In particular, nodes that are not in the ready state and are not newly created (i.e. not having the `node.cloudprovider.kubernetes.io/uninitialized` taint) now get removed from Azure load balancers. ([#109932](https://github.com/kubernetes/kubernetes/pull/109932), [@ricky-rav](https://github.com/ricky-rav)) [SIG Cloud Provider]