* Registries backed by the generic Store's `Update` implementation support delete-on-update, which allows resources to be automatically deleted during an update provided: ([#48065](https://github.com/kubernetes/kubernetes/pull/48065), [@ironcladlou](https://github.com/ironcladlou))
        * Garbage collection is enabled for the Store
        * The resource being updated has no finalizers

* specify azure file share name in azure file plugin ([#76988](https://github.com/kubernetes/kubernetes/pull/76988), [@andyzhangx](https://github.com/andyzhangx))
* [metrics-server addon] Restore connecting to nodes via IP addresses ([#76819](https://github.com/kubernetes/kubernetes/pull/76819), [@serathius](https://github.com/serathius))

  clock desync. ([#118922](https://github.com/kubernetes/kubernetes/pull/118922), [@champtar](https://github.com/champtar))
- `plugin_evaluation_total` metric supports prescore/score extension point.
  The metric doesn't get incremented when the prescore/score plugin has nothing to do with an incoming pod. ([#118025](https://github.com/kubernetes/kubernetes/pull/118025), [@AxeZhan](https://github.com/AxeZhan))

### Documentation

  change, rather than reliably updating every 60 seconds (or whatever). If you
  are trying to monitor for whether iptables updates are failing, the
  `sync_proxy_rules_iptables_restore_failures_total` metric may be more useful. ([#81517](https://github.com/kubernetes/kubernetes/pull/81517), [@danwinship](https://github.com/danwinship))
### Instrumentation

- Restores priority of static control plane pods in the cluster/gce/manifests control-plane manifests ([#89970](https://github.com/kubernetes/kubernetes/pull/89970), [@liggitt](https://github.com/liggitt)) [SIG Cluster Lifecycle and Node]

### Other notable changes

* Kubernetes json deserializer is now case-sensitive to restore compatibility with pre-1.8 servers. ([#65157](https://github.com/kubernetes/kubernetes/pull/65157), [@caesarxuchao](https://github.com/caesarxuchao))

## Dependencies

### Added
- github.com/checkpoint-restore/go-criu/v5: [v5.0.0](https://github.com/checkpoint-restore/go-criu/v5/tree/v5.0.0)
- github.com/frankban/quicktest: [v1.11.3](https://github.com/frankban/quicktest/tree/v1.11.3)

### Changed

    *   Passes sourcesReady from kubelet to the device plugin to avoid removing inactive pods during the grace period of kubelet restart.
    *   Extends the gpu_device_plugin e2e_node test to verify that scheduled pods can continue to run even after a device plugin deletion and kubelet restart.
*   The NodeController no longer supports kubelet 1.2. ([#48996](https://github.com/kubernetes/kubernetes/pull/48996),[ @k82cn](https://github.com/k82cn))

1.6 will also default to protobuf encoding if using etcd3.  **This change is
irreversible.**  To rollback, you must restore from a backup made before the
protobuf/etcd3 switch, and any changes since the backup will be lost.  As 1.5
does not support protobuf encoding, if you roll back to 1.5 after upgrading to
protobuf you will be forced to restore from backup, and you will lose any changes
since you converted to protobuf.  After conversion to protobuf, you should

- The Scheduler did not run the plugin Score method when its PreScore method returned a Skip status. In other words, the PreScore/Score plugin could return a Skip status in PreScore if the plugin did nothing in Score for that Pod. ([#115652](https://github.com/kubernetes/kubernetes/pull/115652), [@AxeZhan](https://github.com/AxeZhan))