## 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

- Incorporating feedback on PR #119341 ([#120087](https://github.com/kubernetes/kubernetes/pull/120087), [@divyasri537](https://github.com/divyasri537)) [SIG API Machinery]

   1. Currently, if users were to explicitly specify CacheSize of 0 for KMS provider, they would end-up with a provider that caches up to 1000 keys. This PR changes this behavior.
      Post this PR, when users supply 0 for CacheSize this will result in a validation error.

		if store.getUsersSysType() == LDAPUsersSysType {
			// Add a fallback removal towards previous content that may come back
			// as a ghost user due to lack of delete, this change occurred
			// introduced in PR #11840
			store.deleteMappedPolicy(ctx, name, regUser, false)
		}
		err := store.deleteMappedPolicy(ctx, name, userType, isGroup)
		if err != nil && !errors.Is(err, errNoSuchPolicy) {
			return updatedAt, err

- This PR adds additional validation for endpoint ip configuration while iterating through queried endpoint list. ([#117225](https://github.com/kubernetes/kubernetes/pull/117225), [@princepereira](https://github.com/princepereira)) [SIG Network and Windows]

- The sidecars' finish time will now be accounted for when calculating the job's finish time. ([#124942](https://github.com/kubernetes/kubernetes/pull/124942), [@AxeZhan](https://github.com/AxeZhan)) [SIG Apps]
- This PR added tracing support to the kubelet's read-only endpoint, which currently
  does not have tracing. It makes use of the `WithPublicEndpoint` option to prevent callers

evict running Pods that do not tolerate the NoSchedule taint if the taint is added to a Node.  There is an open PR ([#48189](https://github.com/kubernetes/kubernetes/pull/48189)) to resolve this issue, but as this issue also exists in 1.6, and as we do not wish to risk release stability by merging it directly prior to a release without sufficient testing, we have decided to defer merging the PR until the next point release for each minor version ([#48190](https://github.com/kubernetes/kubernetes/pull/48190))....

            backwards. See PR
            [47925](https://github.com/tensorflow/tensorflow/pull/47925).
        *   `tf.image.crop_and_resize` gradient w.r.t. either `image` or
            `boxes`. See PR
            [48905](https://github.com/tensorflow/tensorflow/pull/48905).
        *   `tf.sparse.sparse_dense_matmul` forwards. See PR

* Fix --help flag parsing  ([#74682](https://github.com/kubernetes/kubernetes/pull/74682), [@soltysh](https://github.com/soltysh))
* This PR removes the following metrics: ([#74636](https://github.com/kubernetes/kubernetes/pull/74636), [@logicalhan](https://github.com/logicalhan))
    *   reflector_items_per_list
    *   reflector_items_per_watch

- SourcesReady provides the readiness of kubelet configuration sources such as apiserver update readiness. ([#81344](https://github.com/kubernetes/kubernetes/pull/81344), [@zouyee](https://github.com/zouyee))