- CRI: exposed commit memory bytes in container stats specific to Windows ([#119238](https://github.com/kubernetes/kubernetes/pull/119238), [@kiashok](https://github.com/kiashok))
- Client-go now exposes two new metrics to monitor the client-go logic that

    /** The key of the message: Multiple */
    public static final String LABELS_search_result_sort_multiple = "{labels.search_result_sort_multiple}";

    /** The key of the message: {0} bytes */
    public static final String LABELS_search_result_size = "{labels.search_result_size}";

    /** The key of the message: Registered: */

    return valueStrength != Strength.STRONG;
  }

  enum Strength {
    /*
     * TODO(kevinb): If we strongly reference the value and aren't loading, we needn't wrap the
     * value. This could save ~8 bytes per entry.
     */

    STRONG {
      @Override
      <K, V> ValueReference<K, V> referenceValue(
          Segment<K, V> segment, ReferenceEntry<K, V> entry, V value, int weight) {
        return (weight == 1)

pkg archive/tar, var ErrInsecurePath error #55356
pkg archive/zip, var ErrInsecurePath error #55356
pkg bytes, func Clone([]uint8) []uint8 #45038
pkg bytes, func CutPrefix([]uint8, []uint8) ([]uint8, bool) #42537
pkg bytes, func CutSuffix([]uint8, []uint8) ([]uint8, bool) #42537
pkg context, func Cause(Context) error #51365
pkg context, func WithCancelCause(Context) (Context, CancelCauseFunc) #51365
pkg context, type CancelCauseFunc func(error) #51365

- Fixed graceful apiserver shutdown to not drop outgoing bytes before the process terminates. ([#72970](https://github.com/kubernetes/kubernetes/pull/72970), [@sttts](https://github.com/sttts))

### Apps

- The maximum length of the `CSINode` id field has increased to 256 bytes to match the CSI spec. ([#104160](https://github.com/kubernetes/kubernetes/pull/104160), [@pacoxu](https://github.com/pacoxu))

- Client-go: fixed potential data races retrying requests using a custom `io.Reader` body; with this fix, only requests with no body or with `string` / `[]byte` / `runtime.Object` bodies can be retried ([#113933](https://github.com/kubernetes/kubernetes/pull/113933), [@liggitt](https://github.com/liggitt))

*   Reorder some TF-GAN loss functions in a non-backwards compatible way.

## Known Issues

*   In Python 3, `Dataset.from_generator()` does not support Unicode strings.
    You must convert any strings to bytes objects before yielding them from the
    generator.

## Thanks to our Contributors

This release contains contributions from many people at Google, as well as:

- Fixed a bug in the JSON frame reader that could cause it to retain a reference to the underlying array of the byte slice passed to read. ([#123620](https://github.com/kubernetes/kubernetes/pull/123620), [@benluddy](https://github.com/benluddy))

* Add per-pod scheduling metrics across 1 or more schedule attempts. ([#83674](https://github.com/kubernetes/kubernetes/pull/83674), [@liu-cong](https://github.com/liu-cong))
* Fix validation message to mention bytes, not characters. ([#80880](https://github.com/kubernetes/kubernetes/pull/80880), [@DirectXMan12](https://github.com/DirectXMan12))