- `rest_client_request_latency_seconds` -> `rest_client_request_duration_seconds`
      - `scheduler_scheduling_latency_seconds` -> `scheduler_scheduling_duration_seconds `
      - `docker_operations` -> `docker_operations_total`
      - `docker_operations_latency_microseconds` -> `docker_operations_duration_seconds`
      - `docker_operations_errors` -> `docker_operations_errors_total`

- Deprecated metric `rest_client_request_latency_seconds` has been turned off. ([#83836](https://github.com/kubernetes/kubernetes/pull/83836), [@RainbowMango](https://github.com/RainbowMango))
- Following metrics from kubelet are now marked as with the ALPHA stability level:
  `node_cpu_usage_seconds_total`
  `node_memory_working_set_bytes`
  `container_cpu_usage_seconds_total`
  `container_memory_working_set_bytes`
  `scrape_error`

- `kube-proxy` in iptables mode will now have separate `sync_full_proxy_rules_duration_seconds`\nand
  `sync_partial_proxy_rules_duration_seconds` (in addition to the existing\n`sync_proxy_rules_duration_seconds`),
  giving better information about the duration of each \nsync type, rather than

identified in large-scale clusters when deleting thousands of pods simultaneously across hundreds of nodes. Kubelets in this scenario can encounter temporarily increased latency of `delete pod` API calls -- above the target service level objective of 1 second. If you run clusters with this usage pattern and if pod deletion latency could be an issue for you, you might want to wait until the issue is resolved before you upgrade.

For more information and for updates on resolution of this issue,...

    *   reflector_items_per_list
    *   reflector_items_per_watch
    *   reflector_last_resource_version
    *   reflector_list_duration_seconds
    *   reflector_lists_total
    *   reflector_short_watches_total
    *   reflector_watch_duration_seconds
    *   reflector_watches_total

- Kubelet TCP and HTTP probes are now more effective using networking resources:
  conntrack entries, sockets. This is achieved by reducing the `TIME-WAIT` state
  of the connection to 1 second, instead of the defaults 60 seconds. This allows
  kubelet to free the socket, and free conntrack entry and ephemeral port associated. ([#115143](https://github.com/kubernetes/kubernetes/pull/115143), [@aojea](https://github.com/aojea))

- Remove the deprecated metrics "scheduling_algorithm_preemption_evaluation_seconds" and "binding_duration_seconds", suggest to use "scheduler_framework_extension_point_duration_seconds" instead. ([#96447](https://github.com/kubernetes/kubernetes/pull/96447), [@chendave](https://github.com/chendave)) [SIG Cluster Lifecycle, Instrumentation, Scheduling and Testing]

  For more information, see [#57198](https://github.com/kubernetes/kubernetes/issues/57198).

*   Mutating and validating webhook configurations are continuously polled by the API server (once per second). This issue will be fixed in a future release.

  For more information, see [#56357](https://github.com/kubernetes/kubernetes/issues/56357).

* GCE/GKE load balancer health check default interval changes from 2 seconds to 8 seconds, unhealthyThreshold to 3. ([#70099](https://github.com/kubernetes/kubernetes/pull/70099), [@grayluck](https://github.com/grayluck))
    * Health check parameters are configurable to be bigger than default values.

* fix azure disk attachment error on Linux ([#70002](https://github.com/kubernetes/kubernetes/pull/70002), [@andyzhangx](https://github.com/andyzhangx))
* GCE/GKE load balancer health check default interval changes from 2 seconds to 8 seconds, unhealthyThreshold to 3. ([#70099](https://github.com/kubernetes/kubernetes/pull/70099), [@grayluck](https://github.com/grayluck))
    * Health check parameters are configurable to be bigger than default values.