- Mark `net.ipv4.ip_unprivileged_port_start` as safe `sysctl`. ([#103326](https://github.com/kubernetes/kubernetes/pull/103326), [@pacoxu](https://github.com/pacoxu))
- Metrics server nanny has now poll period set to 30s (previously 5 minutes) to allow faster scaling of metrics server. ([#101869](https://github.com/kubernetes/kubernetes/pull/101869), [@olagacek](https://github.com/olagacek)) [SIG Cloud Provider and Instrumentation]

- If `kubeadm reset` finds no etcd member ID for the peer it removes during the `remove-etcd-member` phase, it continues immediately to other phases, instead of retrying the phase for up to 3 minutes before continuing. ([#118192](https://github.com/kubernetes/kubernetes/pull/118192), [@dlipovetsky](https://github.com/dlipovetsky)) [SIG Cluster Lifecycle]

and "unreachable" using `NoExecute` taints. In combination with `tolerationSeconds` described below, this allows per-pod specification of how long to remain bound to a node that becomes unreachable or not ready, rather than using the default of 5 minutes. You can enable this alpha feature by including `TaintBasedEvictions=true` in `--feature-gates`, such as `--feature-gates=FooBar=true,TaintBasedEvictions=true`. Documentation is [here](https://kubernetes.io/docs/concepts/configuration/assign-pod-node/)....

* Fixes bug in DaemonSetController causing it to stop processing some DaemonSets for 5 minutes after node removal. ([#76060](https://github.com/kubernetes/kubernetes/pull/76060), [@krzysztof-jastrzebski](https://github.com/krzysztof-jastrzebski))

##### Cluster Autoscaler

* Cluster autoscaler is now GA
* Cluster support size is increased to 1000 nodes
* Respect graceful pod termination of up to 10 minutes
* Handle zone stock-outs and failures
* Improve monitoring and error reporting

#### Container Runtime Interface (CRI)

- Kubelet: the connection to a DRA driver became unusable because of an internal deadlock when a connection was idle for 30 minutes. ([#133934](https://github.com/kubernetes/kubernetes/pull/133934), [@pohly](https://github.com/pohly)) [SIG Node]

- Kubeadm: fixed the bug where the v1beta4 Timeouts.EtcdAPICall field was not respected in etcd client operations, and the default timeout of 2 minutes was always used. ([#129862](https://github.com/kubernetes/kubernetes/pull/129862), [@neolit123](https://github.com/neolit123)) [SIG Cluster Lifecycle]

### Other (Cleanup or Flake)

        pipeline deterministic. This may lead to increased model loading time,
        because time spent in graph optimizations is now unbounded (was 20
        minutes).
*   Deterministic Op Functionality (enabled by setting `TF_DETERMINISTIC_OPS` to
    `"true"` or `"1"`):
    *   Add a deterministic GPU implementation of
        `tf.nn.softmax_cross_entropy_with_logits`. See PR

pkg time, method (*Time) GobDecode([]uint8) error
pkg time, method (*Time) UnmarshalJSON([]uint8) error
pkg time, method (*Timer) Stop() bool
pkg time, method (Duration) Hours() float64
pkg time, method (Duration) Minutes() float64
pkg time, method (Duration) Nanoseconds() int64
pkg time, method (Duration) Seconds() float64
pkg time, method (Duration) String() string
pkg time, method (Month) String() string