- Added `job_pods_creation_total` metrics for tracking Pods created by the Job controller labeled by events which triggered the Pod creation. ([#121481](https://github.com/kubernetes/kubernetes/pull/121481), [@dejanzele](https://github.com/dejanzele))
- Added `kubectl node drain` helper callbacks `OnPodDeletionOrEvictionStarted`

- Fixed a bug where a change in the `appProtocol` for a Service did not trigger a load balancer update. ([#112785](https://github.com/kubernetes/kubernetes/pull/112785), [@MartinForReal](https://github.com/MartinForReal)) [SIG Cloud Provider and Network]

- In some cases, a Flex volume resize may leave a PVC with erroneous Resizing condition even after volume has...

- `(*"k8s.io/client-go/rest".Request).{Do,DoRaw,Stream,Watch}` now require callers to pass a `context.Context` as an argument. The context is used for timeout and cancellation signaling and to pass supplementary information to round trippers in the wrapped transport chain. If you don't need any of this functionality, it is sufficient to pass a context created with `context.Background()` to these functions. The `(*"k8s.io/client-go/rest".Request).Context` method is removed now that all...

  * Marking a StorageClass as default makes it so that even a PersistentVolumeClaim without a StorageClass specified will trigger dynamic provisioning (instead of binding to an existing pool of PVs).

* fix a bug where disk pressure could trigger prematurely when using overlay2 ([#53684](https://github.com/kubernetes/kubernetes/pull/53684), [@dashpole](https://github.com/dashpole))

- Fixed a bug where a change in the `appProtocol` for a Service did not trigger a load balancer update. ([#113031](https://github.com/kubernetes/kubernetes/pull/113031), [@MartinForReal](https://github.com/MartinForReal)) [SIG Cloud Provider and Network]

would be a pod set to `Failed` phase with reason set to `OutOfCpu` or `OutOfMemory`, but any resource on the node that has some fixed limit (including persistent volume counts on cloud nodes, exclusive CPU cores, or unique hardware devices) could trigger the failure. While this behavior is correct it reduces the throughput of pod execution and creates user-visible warnings - [future versions of Kubernetes will minimize the likelihood users see pod failures due to this issue](https://github.com/ku...

    *   `converter.convert()` no longer returns a `tf.function`. Now the
        function must be accessed from the saved model.
    *   The `converter.calibrate()` method has been removed. To trigger
        calibration, a `calibration_input_fn` should be provided to
        `converter.convert()`.

*   Other:

    *   Fix accidental quadratic graph construction cost in graph-mode