- Ensures `ExecProbeTimeout=false` kubelet feature gate with dockershim is taken into account, when the exec probe takes longer than `timeoutSeconds` configuration. ([#100200](https://github.com/kubernetes/kubernetes/pull/100200), [@jackfrancis](https://github.com/jackfrancis))

*   Pod preemption has been improved.
    *   Now takes PodDisruptionBudget into account. ([#56178](https://github.com/kubernetes/kubernetes/pull/56178),[ @bsalamat](https://github.com/bsalamat))

- Implemented `event_handling_duration_seconds` metric, which is the time the scheduler takes to handle each kind of events. ([#125929](https://github.com/kubernetes/kubernetes/pull/125929), [@sanposhiho](https://github.com/sanposhiho))
- Implemented `queueing_hint_execution_duration_seconds` metric, which is the time the QueueingHint function takes. ([#126227](https://github.com/kubernetes/kubernetes/pull/126227), [@sanposhiho](https://github.com/sanposhiho))

job controller back-off logic is now decoupled from workqueue. In case of parallelism > 1, if there are multiple new failures in a reconciliation cycle, all the failures are taken into account to compute the back-off. Previously, the back-off kicked in for all types of failures; with this change, only pod failures are taken into account. If the back-off limits exceeds, the job is marked as failed immediately; before this change, the job is marked as failed in the next back-off. ([#114768](https:...

- The kube-apiserver's healthz now takes in an optional query parameter which allows you to disable health checks from causing healthz failures.  ([#70676](https://github.com/kubernetes/kubernetes/pull/70676), [@logicalhan](https://github.com/logicalhan))

* Fix a potential bug that scheduler preempts unnecessary pods. ([#70898](https://github.com/kubernetes/kubernetes/pull/70898), [@Huang-Wei](https://github.com/Huang-Wei))
* The kube-apiserver's healthz now takes in an optional query parameter which allows you to disable health checks from causing healthz failures.  ([#70676](https://github.com/kubernetes/kubernetes/pull/70676), [@logicalhan](https://github.com/logicalhan))

- Kubeadm now makes sure the etcd manifest is regenerated upon upgrade even when no etcd version change takes place ([#94395](https://github.com/kubernetes/kubernetes/pull/94395), [@rosti](https://github.com/rosti)) [SIG Cluster Lifecycle]

- The apimachinery util/net function used to detect the bind address `ResolveBindAddress()` takes into consideration global IP addresses on loopback interfaces when 1) the host has default routes, or 2) there are no global IPs on those interfaces in order to support more complex network scenarios like BGP Unnumbered RFC 5549 ([#95790](https://...

- A new kubelet metric `image_pull_duration_seconds` was added. The metric tracks the duration (in seconds) it takes for an image to be pulled, including the time spent in the waiting queue of image puller. The metric is broken down by bucketed image size. ([#121719](https://github.com/kubernetes/kubernetes/pull/121719), [@ruiwen-zhao](https://github.com/ruiwen-zhao))...

econds metrics because they were reporting seconds rather than milliseconds. They were also subject to multiple naming guideline violations (units should be in base units and "duration" is the best practice labelling to measure the time a request takes). Please convert to use *_admission_duration_seconds and *_admission_duration_seconds_summary, as these now report the unit as described, and follow the instrumentation best practices. ([#75279](https://github.com/kubernetes/kubernetes/pull/75279),...