- Fixed a race condition that could result in erroneous volume unmounts for flex volume plugins during kubelet restart. ([#127669](https://github.com/kubernetes/kubernetes/pull/127669), [@olyazavr](https://github.com/olyazavr))

- It introduces a new scope name `VolumeAttributesClass`. 
  
  It matches all PVC objects that have the volume attributes class mentioned. 
  

- Kubelet: Fix the volume manager didn't check the device mount state in the actual state of the world before marking the volume as detached. It may cause a pod to be stuck in the Terminating state due to the above issue when it was deleted. ([#129063](https://github.com/kubernetes/kubernetes/pull/129063), [@carlory](https...

 - Kubelet flag `--keep-terminated-pod-volumes` was removed.  This flag was deprecated in 2017. ([#122082](https://github.com/kubernetes/kubernetes/pull/122082), [@carlory](https://github.com/carlory)) [SIG Apps, Node, Storage and Testing]
 
## Changes by Kind

### Deprecation

- Enabled kube-apiserver support for `PodCertificateRequest` and `PodCertificate` projected volumes (behind the `PodCertificateRequest` feature gate). ([#128010](https://github.com/kubernetes/kubernetes/pull/128010), [@ahmedtd](https://github.com/ahmedtd)) [SIG API Machinery, Apps, Auth, Cloud Provider, Etcd, Node, Storage and Testing]

 *
 * <p>The views returned by {@link #column}, {@link #columnKeySet()}, and {@link #columnMap()} have
 * iterators that don't support {@code remove()}. Otherwise, all optional operations are supported.
 * Null row keys, columns keys, and values are not supported.
 *
 * <p>Lookups by row key are often faster than lookups by column key, because the data is stored in

 *
 * <p>The views returned by {@link #column}, {@link #columnKeySet()}, and {@link #columnMap()} have
 * iterators that don't support {@code remove()}. Otherwise, all optional operations are supported.
 * Null row keys, columns keys, and values are not supported.
 *
 * <p>Lookups by row key are often faster than lookups by column key, because the data is stored in

   * changes the contents of the transposed table and vice versa.
   *
   * <p>The returned table supports update operations as long as the input table supports the
   * analogous operation with swapped rows and columns. For example, in a {@link HashBasedTable}
   * instance, {@code rowKeySet().iterator()} supports {@code remove()} but {@code
   * columnKeySet().iterator()} doesn't. With a transposed {@link HashBasedTable}, it's the other

  /**
   * {@inheritDoc}
   *
   * <p>The returned set has an iterator that does not support {@code remove()}.
   *
   * <p>The set's iterator traverses the columns of the first row, the columns of the second row,
   * etc., skipping any columns that have appeared previously.
   */
  @Override
  public Set<C> columnKeySet() {
    Set<C> result = columnKeySet;

 *
 * <p><b>Warning:</b> {@code ArrayTable} is rarely the {@link Table} implementation you want. First,
 * it requires that the complete universe of rows and columns be specified at construction time.
 * Second, it is always backed by an array large enough to hold a value for every possible
 * combination of row and column keys. (This is rarely optimal unless the table is extremely dense.)