* Improvements and fixes in Keras loss masking:
    * Whether you represent a ragged tensor as a `tf.RaggedTensor` or using [keras masking](https://www.tensorflow.org/guide/keras/masking_and_padding), the returned loss values should be the identical to each other. In previous versions Keras may have silently ignored the mask.

### Holding multiple locks

You're allowed to take the Http2Connection lock while holding the Http2Writer lock. But not vice-versa. Because taking the Http2Writer lock can block.

    when (this) {
        is SyntheticJavaPropertyDescriptor -> return KtSymbolOrigin.JAVA_SYNTHETIC_PROPERTY
        is SyntheticFieldDescriptor -> return KtSymbolOrigin.PROPERTY_BACKING_FIELD
        is SamConstructorDescriptor -> return KtSymbolOrigin.SAM_CONSTRUCTOR
        is JavaClassDescriptor, is JavaCallableMemberDescriptor -> return KtSymbolOrigin.JAVA

     *
     * We need to apply the settings and ack them atomically. This is because some HTTP/2
     * implementations (nghttp2) forbid peers from taking advantage of settings before they have
     * acknowledged! In particular, we shouldn't send frames that assume a new `initialWindowSize`
     * until we send the frame that acknowledges this new size.
     *

      }

      return toString()
    }
  }

  fun newBuilder(): Builder = Builder(this)

  /**
   * Builds a cookie. The [name], [value], and [domain] values must all be set before calling
   * [build].
   */
  class Builder() {
    private var name: String? = null
    private var value: String? = null
    private var expiresAt = MAX_DATE
    private var domain: String? = null

   ([#103523](https://github.com/kubernetes/kubernetes/pull/103523), [@divyenpatel](https://github.com/divyenpatel))
- `MaxUnavailable` for `StatefulSets`, allows faster `RollingUpdate` by taking down more than 1 pod at a time.
  The number of pods you want to take down during a `RollingUpdate` is configurable using `maxUnavailable` parameter.

    - [KubeSchedulerConfiguration graduates to Beta](#kubeschedulerconfiguration-graduates-to-beta)
    - [CSI Migration - AzureDisk and vSphere (beta)](#csi-migration---azuredisk-and-vsphere-beta)
    - [Storage capacity tracking](#storage-capacity-tracking)
    - [CSI Volume health monitoring](#csi-volume-health-monitoring)
    - [General ephemeral volumes](#general-ephemeral-volumes)

### Failing Test

- Kubelet: the HostPort implementation in dockershim was not taking into consideration the HostIP field, causing that the same HostPort can not be used with different IP addresses.

  Jobs that are tracked using this mechanism get the annotation `batch.kubernetes.io/job-tracking`. This is a temporary measure. Two releases after this feature graduates to beta, the annotation won't be added to Jobs anymore. ([#98817](https://github.com/kubernetes/kubernetes/pull/98817), ...

- For raw block CSI volumes on Kubernetes, kubelet was incorrectly calling CSI NodeStageVolume for every single "map" (i.e. raw block "mount") operation for a volume already attached to the node. This PR ensures it is only called once per volume per node. ([#112403](https://github.com/kubernetes/kubernetes/pull/112403),...