* Admit critical pods in the kubelet ([#38836](https://github.com/kubernetes/kubernetes/pull/38836), [@bprashanth](https://github.com/bprashanth))
* assign -998 as the oom_score_adj for critical pods (e.g. kube-proxy) ([#39114](https://github.com/kubernetes/kubernetes/pull/39114), [@dchen1107](https://github.com/dchen1107))

- There is now support for collecting pod logs under `/var/log/pods/NAMESPACE_NAME_UID` to stackdriver with `k8s_pod` resource type. ([#74502](https://github.com/kubernetes/kubernetes/pull/74502), [@Random-Liu](https://github.com/Random-Liu))

- When using `kubectl logs <pod-name>` and the pod is not found, the error message now includes the namespace. Previously, the message would be "Error from server (NotFound): pods "my-pod-name" not found". Now, it reflects the namespace in the message as follows: "Error from server (NotFound): pods "my-pod-name" not found in namespace "default"".

*   StatefulSet controller now creates a label for each Pod in a StatefulSet. The label is `statefulset.kubernetes.io/pod-name`, where `pod-name` = the name of the Pod. This allows users to create a Service per Pod to expose a connection to individual Pods. ([#55329](https://github.com/kubernetes/kubernetes/pull/55329),[ @kow3ns](https://github.com/kow3ns))

- The calculation of pod UIDs for static pods has changed to ensure each static pod gets a unique value - this will cause all static pod containers to be recreated/restarted if an in-place kubelet upgrade from 1.20 to 1.21 is performed. Note that draining pods before upgrading the kubelet across minor versions is the supported upgrade path. ([#87461](https://github.com/kubernetes/kubernetes/pull/87461), [@bboreham](https://github.com/bboreham))...

* **[Pod priority and preemption](https://kubernetes.io/docs/concepts/configuration/pod-priority-preemption/) is now enabled by default.** Note that this means that pods from *any* namespace can now request priority classes that compete with and/or cause preemption of critical system pods that are already running. If that is not desired, disable the PodPriority feature by setting `--feature-gat...

versions. There are still two notable changes: 1. TaintBasedEviction is moved to Beta and will be enabled by default. With this feature enabled, condition taints are automatically added to the nodes and pods can add tolerations for them if needed. 2. Pod critical annotation is deprecated. Pods should use pod priority instead of the annotation.

It is worth noting again that kube-scheduler will use apiVersion `kubescheduler.config.k8s.io/v1alpha1` instead of `componentconfig/v1alpha1` in its...

  the Job controller now does not consider a terminating Pod (a pod that has a `.metadata.deletionTimestamp`)
  as a failure until that Pod is terminal (its `.status.phase` is `Failed` or `Succeeded`).
  
  However, the Job controller creates a replacement Pod as soon as the termination becomes apparent.
  Once the pod terminates, the Job controller evaluates `.backoffLimit` and `.podFailurePolicy` for the

- Fix kubelet fail to delete orphaned pod directory when the kubelet's pods directory (default is `/var/lib/kubelet/pods`) symbolically links to another disk device's directory ([#79094](https://github.com/kubernetes/kubernetes/pull/79094), [@gaorong](https://github.com/gaorong))

## Testing

* kubectl port-forward now allows using resource name (e.g., deployment/www) to select a matching pod, as well as the use of --pod-running-timeout to wait until at least one pod is running. ([#59705](https://github.com/kubernetes/kubernetes/pull/59705), [@phsiao](https://github.com/phsiao))