* purposes, the "count" field, tracking the number of elements, serves as that volatile
     * variable ensuring visibility. This is convenient because this field needs to be read in many
     * read operations anyway:
     *
     * - All (unsynchronized) read operations must first read the "count" field, and should not
     * look at table entries if it is 0.
     *

    memory!

    The best practice in OkHttp 3 is to create a single OkHttpClient instance
    and share it throughout the application. Requests that needs a customized
    client should call `OkHttpClient.newBuilder()` on that shared instance.
    This allows customization without the drawbacks of separate connection
    pools.

#### Storage
- A node that uses a CSI raw block volume needs to be drained before kubelet can be upgraded to 1.17. ([#74026](https://github.com/kubernetes/kubernetes/pull/74026), [@mkimuram](https://github.com/mkimuram))

#### Windows

     * The key reference queue contains entries whose keys have been garbage collected, and which
     * need to be cleaned up internally.
     */
    final @Nullable ReferenceQueue<K> keyReferenceQueue;

    /**
     * The value reference queue contains value references whose values have been garbage collected,
     * and which need to be cleaned up internally.
     */
    final @Nullable ReferenceQueue<V> valueReferenceQueue;

- kube-controller-manager currently needs a writable `--cert-dir` (default is `/var/run/kubernetes`) for generating self-signed certificates, when no `--tls-cert-file` or `--tls-private-key-file` are provided.

     * The key reference queue contains entries whose keys have been garbage collected, and which
     * need to be cleaned up internally.
     */
    final @Nullable ReferenceQueue<K> keyReferenceQueue;

    /**
     * The value reference queue contains value references whose values have been garbage collected,
     * and which need to be cleaned up internally.
     */
    final @Nullable ReferenceQueue<V> valueReferenceQueue;

* Action required: validation rule on metadata.initializers.pending[x].name is tightened. The initializer name needs to contain at least three segments separated by dots. If you create objects with pending initializers, (i.e., not relying on apiserver adding pending initializers according to initializerconfiguration), you need to update the initializer name in existing objects and in configuration files to comply to the new validation rule. ([#51283](https://github....

     * purposes, the "count" field, tracking the number of elements, serves as that volatile
     * variable ensuring visibility. This is convenient because this field needs to be read in many
     * read operations anyway:
     *
     * - All (unsynchronized) read operations must first read the "count" field, and should not
     * look at table entries if it is 0.
     *

* The deprecated kubelet --configure-cbr0 flag has been removed, and with that the "classic" networking mode as well.  If you depend on this mode, please investigate whether the other network plugins `kubenet` or `cni` meet your needs. ([#34906](https://github.com/kubernetes/kubernetes/pull/34906), [@luxas](https://github.com/luxas))

- When using `kubectl replace` (or the equivalent API call) on a Service, the caller no longer needs to do a read-modify-write cycle to fetch the allocated values for `.spec.clusterIP` and `.spec.ports[].nodePort`.  Instead the API server will automatically carry these forward from the original object when the new object does not specify them. ([...