* Fixes a performance issue ([#51899](https://github.com/kubernetes/kubernetes/pull/51899)) identified in large-scale clusters when deleting thousands of pods simultaneously across hundreds of nodes, by actively removing containers of deleted pods, rather than waiting for periodic garbage collection and batching resulting pod API deletion requests. ([#53233](https://github.com/kubernetes/kubernetes/pull/53233), [@dashpole](https://...

##### **Other**

- Introduction of the pod overhead feature to the scheduler. This functionality is alpha-level as of
  Kubernetes v1.16, and is only honored by servers that enable the PodOverhead feature.gate. ([#78319](https://github.com/kubernetes/kubernetes/pull/78319), [@egernst](https://github.com/egernst))

## Changes by Kind

### Bug or Regression

- Fix an ephemeral port exhaustion bug caused by improper connection management that occurred when a large number of objects were handled by kubectl while exec auth was in use. ([#112339](https://github.com/kubernetes/kubernetes/pull/112339), [@enj](https://github.com/enj)) [SIG API Machinery and Auth]

can now register for a single version of a resource (for example, `apps/v1 deployments`) and be called when any other version of that resource is modified (for example `extensions/v1beta1 deployments`). This allows new versions of a resource to be handled by admission webhooks without needing to update every webhook to understand the new version. See the API documentation for the `matchPolicy: Equivalent` option in MutatingWebhookConfiguration and ValidatingWebhookConfiguration types. ([#78135](ht...

- Disabled `LocalStorageCapacityIsolation` feature gate is honored during scheduling. ([#96181](https://github.com/kubernetes/kubernetes/pull/96181), [@Huang-Wei](https://github.com/Huang-Wei)) [SIG Scheduling]

object. This could disrupt the ability of clients clients to re-establish watches properly. ([#58547](https://github.com/kubernetes/kubernetes/pull/58547), [@liggitt](https://github.com/liggitt))

* kube-apiserver: requests to endpoints handled by unavailable extension API servers (as indicated by an `Available` condition of `false` in the registered APIService) now return `503` errors instead of `404` errors. ([#58070](https://github.com/kubernetes/kubernetes/pull/58070), [@weekface](h...

- The `--leader-elect*` CLI args are now honored in scheduler. ([#105915](https://github.com/kubernetes/kubernetes/pull/105915), [@Huang-Wei](https://github.com/Huang-Wei))
- The `--leader-elect*` CLI args are now honored in the scheduler. ([#105712](https://github.com/kubernetes/kubernetes/pull/105712), [@Huang-Wei](https://github.com/Huang-Wei))

* Fixed controller manager crash by making it tolerant to discovery errors.([#49767](https://github.com/kubernetes/kubernetes/pull/49767), [@deads2k](https://github.com/deads2k))

- Added `.spec.completionMode` field to Job, with accepted values `NonIndexed` (default) and `Indexed`. This is an alpha field and is only honored by servers with the `IndexedJob` feature gate enabled. ([#98441](https://github.com/kubernetes/kubernetes/pull/98441), [@alculquicondor](https://github.com/alculquicondor)) [SIG Apps and CLI]