- kube-scheduler

  - Migrate scheduler to use v1beta1 Event API. any tool targeting scheduler events needs to use v1beta1 Event API ([#78447](https://github.com/kubernetes/kubernetes/pull/78447), [@yastij](https://github.com/yastij))

- kubeadm

## Changelog since v1.19.15

## Changes by Kind

### Bug or Regression

- Fix detach from non-existant nodes and support dangling volumes (#104912, @gnufied) [SIG Cloud Provider and Storage]
- Revert PR #102925 which introduced unexpected scheduling behavior based on balanced resource allocation (#105526, @harisund) [SIG Scheduling]

## Dependencies

### Added
_Nothing has changed._

### Changed

  - `priority_level_seat_count_watermarks` is removed.
  - `priority_level_request_count_samples` is replaced with `priority_level_request_utilization`, which samples every nanosecond rather than every millisecond; the old metric conveyed utilization despite its name.

* kubeadm: Fix join preflight check false negative ([#49825](https://github.com/kubernetes/kubernetes/pull/49825), [@erhudy](https://github.com/erhudy))
* Revert deprecation of vCenter port in vSphere Cloud Provider. ([#49689](https://github.com/kubernetes/kubernetes/pull/49689), [@divyenpatel](https://github.com/divyenpatel))

* kube-proxy will emit "FailedToStartNodeHealthcheck" event when fails to start healthz server. ([#49267](https://github.com/kubernetes/kubernetes/pull/49267), [@MrHohn](https://github.com/MrHohn))

  `What IS the fun?' said Alice.

  `Why, SHE,' said the Gryphon.  `It's all her fancy, that:  they
never executes nobody, you know.  Come on!'

  `Everybody says "come on!" here,' thought Alice, as she went
slowly after it:  `I never was so ordered about in all my life,
never!'

  They had not gone far before they saw the Mock Turtle in the
distance, sitting sad and lonely on a little ledge of rock, and,

  private static final ListenerCallQueue.Event<Listener> HEALTHY_EVENT =
      new ListenerCallQueue.Event<Listener>() {
        @Override
        public void call(Listener listener) {
          listener.healthy();
        }

        @Override
        public String toString() {
          return "healthy()";
        }
      };
  private static final ListenerCallQueue.Event<Listener> STOPPED_EVENT =

### Other (Cleanup or Flake)

- When retrieving event resources, the reportingController and reportingInstance fields in the event will contain values. ([#120066](https://github.com/kubernetes/kubernetes/pull/120066), [@HirazawaUi](https://github.com/HirazawaUi)) [SIG Instrumentation]

## Dependencies

### Added

So we have a dedicated thread for every socket that just reads frames and dispatches them.

The reader thread must never run application-layer code. Otherwise one slow stream can hold up the entire connection.

 *
 *  * When an entry is being **edited**, it is not necessary to supply data for every value; values
 *    default to their previous value.
 *
 * Every [edit] call must be matched by a call to [Editor.commit] or [Editor.abort]. Committing is
 * atomic: a read observes the full set of values as they were before or after the commit, but never
 * a mix of values.
 *