- Fix clearing of rate-limiter for the queue of checks for cleaning stale pod disruption conditions. 
  The bug could result in the PDB synchronization updates firing too often or the pod disruption cleanups taking too long to happen. ([#114780](https://github.com/kubernetes/kubernetes/pull/114780), [@mimowo](https://github.com/mimowo)) [SIG Apps]

* Pod log attempts are now reported in apiserver prometheus metrics with verb `CONNECT` since they can run for very long periods of time. ([#50123](https://github.com/kubernetes/kubernetes/pull/50123), [@WIZARD-CXY](https://github.com/WIZARD-CXY))

  // If this probe fails, the Pod will be restarted, just as if the livenessProbe failed.
  // This can be used to provide different probe parameters at the beginning of a Pod's lifecycle,
  // when it might take a long time to load data or warm a cache, than during steady-state operation.
  // This cannot be updated.
  // More info: https://kubernetes.io/docs/concepts/workloads/pods/pod-lifecycle#container-probes
  // +optional

  is beta but disabled by default. ([#120198](https://github.com/kubernetes/kubernetes/pull/120198), [@pacoxu](https://github.com/pacoxu))
- `kubectl` will not support the `/swagger-2.0.0.pb-v1` endpoint that has been long deprecated. ([#119410](https://github.com/kubernetes/kubernetes/pull/119410), [@Jefftree](https://github.com/Jefftree))
- `scheduler`: handling of unschedulable pods because a `ResourceClass` is missing

	// `<volume name>` is the name from the `PodSpec.Volumes` array
	// entry. Pod validation will reject the pod if the concatenated name
	// is not valid for a PVC (for example, too long).
	//
	// An existing PVC with that name that is not owned by the pod
	// will *not* be used for the pod to avoid using an unrelated
	// volume by mistake. Starting the pod is then blocked until

	}
	assert.NotContains(t, fp.iptablesData.String(), "-X ", "iptables data unexpectedly contains chain deletions")

	// But resyncing after a long-enough delay will delete the stale chains
	fp.lastIPTablesCleanup = time.Now().Add(-fp.syncPeriod).Add(-1)
	fp.syncProxyRules()

      // The mean, variance, and reserved space outputs of the batch norm op are
      // not used for inference. It doesn't matter what values we provide for
      // the last 5 results as long as they are of the same type. Forward
      // input mean and variance to output mean, variance, reserved_space_1 and
      // reserved_space_2.
      if (std::is_same<FusedBatchNormOpT, TF::FusedBatchNormV2Op>::value) {

	// `<volume name>` is the name from the `PodSpec.Volumes` array
	// entry. Pod validation will reject the pod if the concatenated name
	// is not valid for a PVC (for example, too long).
	//
	// An existing PVC with that name that is not owned by the pod
	// will *not* be used for the pod to avoid using an unrelated
	// volume by mistake. Starting the pod is then blocked until

- Fixed a bug where startupProbe stopped working after a container's first restart ([#101093](https://github.com/kubernetes/kubernetes/pull/101093), [@wzshiming](https://github.com/wzshiming)) [SIG Node]
- Fixed port-forward memory leak for long-running and heavily used connections. ([#99839](https://github.com/kubernetes/kubernetes/pull/99839), [@saschagrunert](https://github.com/saschagrunert)) [SIG API Machinery and Node]

  and `sync_partial_proxy_rules_duration_seconds` (in addition to the existing
  `sync_proxy_rules_duration_seconds`), to give better information about how long
  each sync type is taking, rather than only giving a weighted average of the two
  sync types together. ([#117787](https://github.com/kubernetes/kubernetes/pull/117787), [@danwinship](https://github.com/danwinship)) [SIG Network]