- `requests.deviceclass.resource.k8s.io/<deviceclass>` is charged based on the worst-case number of devices requested.

 that does not depend on mutable selectors.

 - Extended Resources are fully-qualified resource names outside the
 `kubernetes.io` domain. Extended Resource quantities must be integers.
 You can specify any resource name of the form `[aaa.]my-domain.bbb/ccc`

### kubectl (CLI)
* `kubectl create role` and  `kubectl create clusterrole`  invocations must be updated to specify multiple resource names as repeated  `--resource-name` arguments instead of comma-separated arguments to a single `--resource-name` argument. E.g. `--resource-name=x,y` must become `--resource-name x --resource-name y` ([#44950](https://github.com/kubernetes/kubernetes/pull/44950), [@xilabao](https://github.com/xilabao))

- The resource version option, when passed to a list call, is now consistently interpreted as the minimum allowed resource version.  Previously when listing resources that had the watch cache disabled clients could retrieve a snapshot at that exact resource version.  If the client requests a resource version newer than the current state, a TimeoutError is returned suggesting the client retry in a few seconds....

- resource metrics endpoint `/metrics/resource/v1alpha1` as well as all metrics under this endpoint have been deprecated. Please convert to the following metrics emitted by endpoint `/metrics/resource`:
  - scrape_error --> scrape_error
  - node_cpu_usage_seconds_total --> node_cpu_usage_seconds

  For metrics `etcd_request_duration_seconds`, `etcd_requests_total` and `etcd_request_errors_total` replace `type` label with API `resource` and `group` label.
  For metric `apiserver_selfrequest_total` add a API `group` label.
  For metrics `apiserver_watch_events_sizes` and `apiserver_watch_events_total` replace API `kind` label with `resource` label.

* Kubelet will now set extended resource capacity to zero after it restarts. If the extended resource is exported by a device plugin, its capacity will change to a valid value after the device plugin re-connects with the Kubelet. If the extended resource is exported by an external component through direct node status capacity patching, the component should repatch the field after kubelet becomes...

### CVE-2025-4563: Nodes can bypass dynamic resource allocation authorization checks

*   If you run `kubectl get po` while the API server in unreachable, a misleading error is returned: `the server doesn't have a resource type "po"`. To work around this issue, specify the full resource name in the command instead of the abbreviation: `kubectl get pods`. This issue will be fixed in a future release.

* Allow drain --force to remove pods whose managing resource is deleted. ([#41864](https://github.com/kubernetes/kubernetes/pull/41864), [@marun](https://github.com/marun))