*   <li>Return a {@code ClosingFuture}. To turn a {@link ListenableFuture} into a {@code
   *       ClosingFuture} call {@link #from(ListenableFuture)}.
   *   <li>In case this step doesn't create new closeables, you can adapt an API that returns a
   *       {@link ListenableFuture} to return a {@code ClosingFuture} by wrapping it with a call to
   *       {@link #withoutCloser(AsyncFunction)}
   * </ul>
   *
   * <p>Example usage:
   *

 * AsyncCacheLoader.asyncLoad}</a> must return a {@code CompletableFuture}. Users of Guava's {@link
 * com.google.common.util.concurrent.ListenableFuture} can adapt between the two {@code Future}
 * types by using <a href="https://github.com/lukas-krecan/future-converter#java8-guava">{@code
 * net.javacrumbs.futureconverter.java8guava.FutureConverter}</a>.
 *

intensive tests and you should leave some slack for JVM-internal threads like
the garbage collector. And you have to have enough RAM to handle each JVM.

=== Test compatibility.

It is possible to provide a version that allows to adapt the tests behaviour
to older features or bugs that have been changed or fixed in the meantime.

-----------------------------------------
./gradlew test -Dtests.compatibility=1.0.0

* kubelet: add RSS memory to the summary API ([#24015](https://github.com/kubernetes/kubernetes/pull/24015), [@yujuhong](https://github.com/yujuhong))
* e2e: adapt kubelet_perf.go to use the new summary metrics API ([#24003](https://github.com/kubernetes/kubernetes/pull/24003), [@yujuhong](https://github.com/yujuhong))

          int numeratorBits = nBits;
          // This is an upper bound on log2(numerator, ceiling).

          /*
           * We want to do this in long math for speed, but want to avoid overflow. We adapt the
           * technique previously used by BigIntegerMath: maintain separate numerator and
           * denominator accumulators, multiplying the fraction into result when near overflow.
           */

 * AsyncCacheLoader.asyncLoad}</a> must return a {@code CompletableFuture}. Users of Guava's {@link
 * com.google.common.util.concurrent.ListenableFuture} can adapt between the two {@code Future}
 * types by using <a href="https://github.com/lukas-krecan/future-converter#java8-guava">{@code
 * net.javacrumbs.futureconverter.java8guava.FutureConverter}</a>.
 *

* Adding nodeports services to quota ([#22154](https://github.com/kubernetes/kubernetes/pull/22154), [@sdminonne](https://github.com/sdminonne))
* e2e: adapt kubelet_perf.go to use the new summary metrics API ([#24003](https://github.com/kubernetes/kubernetes/pull/24003), [@yujuhong](https://github.com/yujuhong))

  - Make "kubeadm upgrade apply" add the "node-role.kubernetes.io/control-plane" label on existing nodes that only have the "node-role.kubernetes.io/master" label during upgrade.
  - Please adapt your tooling built on top of kubeadm to use the "node-role.kubernetes.io/control-plane" label.

instructed to preemptively tolerate the new taint. For clusters that are being upgraded to 1.24 with `kubeadm upgrade apply`, the command will add the new taint `node-role.kubernetes.io/control-plane:NoSchedule` to existing control plane nodes. Please adapt your infrastructure to these changes. In 1.25 the old taint `node-role.kubernetes.io/master:NoSchedule` will be removed. ([#107533](https://github.com/kubernetes/kubernetes/pull/107533), [@neolit123](https://github.com/neolit123))

## Changes...

        *   `CategoryEncoding` no longer supports `adapt`, `IntegerLookup` now
            supports equivalent functionality. `max_tokens` argument renamed to
            `num_tokens`.
        *   `Discretization` added `num_bins` argument for learning bins
            boundaries through calling `adapt` on a dataset. Renamed `bins`
            argument to `bin_boundaries` for specifying bins without `adapt`.
    *   Improvements to model saving/loading: