* detailed at {@link ImmutableCollection}.
 *
 * <p><b>Warning:</b> as with any sorted collection, you are strongly advised not to use a {@link
 * Comparator} or {@link Comparable} type whose comparison behavior is <i>inconsistent with
 * equals</i>. That is, {@code a.compareTo(b)} or {@code comparator.compare(a, b)} should equal zero
 * <i>if and only if</i> {@code a.equals(b)}. If this advice is not followed, the resulting map will

   - In version 1.8, enable this behavior by setting the
     `--horizontal-pod-autoscaler-use-rest-clients` flag to `true`.

   - In version 1.9, this behavior will be enabled by default, and must be explicitly
     disabled by setting the `--horizontal-pod-autoscaler-use-rest-clients` flag to `false`.

### Cluster Lifecycle

- The `auto-detect` behavior of the kubelet's `--cloud-provider` flag is deprecated.

* fix device name change issue for azure disk ([#60346](https://github.com/kubernetes/kubernetes/pull/60346), [@andyzhangx](https:/...

  /**
   * Creates a fluent iterable with the first {@code size} elements of this fluent iterable. If this
   * fluent iterable does not contain that many elements, the returned fluent iterable will have the
   * same behavior as this fluent iterable. The returned fluent iterable's iterator supports {@code
   * remove()} if this fluent iterable's iterator does.
   *
   * <p><b>{@code Stream} equivalent:</b> {@link Stream#limit} (same).
   *

   * written to or read from it. For example, whether {@code list.get(0) == list.get(0)} is true for
   * the returned list is unspecified.
   *
   * <p>The returned list may have unexpected behavior if it contains {@code NaN}, or if {@code NaN}
   * is used as a parameter to any of its methods.
   *
   * <p>The returned list is serializable.
   *
   * @param backingArray the array to back the list

* Fix GPU resource validation ([#28743](https://github.com/kubernetes/kubernetes/pull/28743), [@therc](https://github.com/therc))
* Scale kube-proxy conntrack limits by cores (new default behavior) ([#28876](https://github.com/kubernetes/kubernetes/pull/28876), [@thockin](https://github.com/thockin))

    visibility = ["//visibility:public"],
)

# By default, XLA GPU is compiled into tensorflow when building with
# --config=cuda even when `with_xla_support` is false. The config setting
# here allows us to override the behavior if needed.
config_setting(
    name = "no_xla_deps_in_cuda",
    define_values = {"no_xla_deps_in_cuda": "true"},
    visibility = ["//visibility:public"],
)

     */
    // NB: yes, this is surprisingly high, but that's what the experiments said was necessary
    // Raising this number slows the worst-case contains behavior, speeds up hashFloodingDetected,
    // and reduces the false-positive probability.
    static final int MAX_RUN_MULTIPLIER = 13;

    /**

- The alpha `Initializers` admission plugin is no longer enabled by default. This matches the off-by-default behavior of the alpha API which drives initializer behavior. ([#66039](https://github.com/kubernetes/kubernetes/pull/66039), [@liggitt](https://github.com/liggitt))

  * Marking a StorageClass as default makes it so that even a PersistentVolumeClaim without a StorageClass specified will trigger dynamic provisioning (instead of binding to an existing pool of PVs).