- [migration phase 1] PodMatchNodeSelector/NodAffinity as filter plugin ([#83660](https://github.com/kubernetes/kubernetes/pull/83660), [@wgliang](https://github.com/wgliang))
- Add more tracing steps in generic_scheduler ([#83539](https://github.com/kubernetes/kubernetes/pull/83539), [@wgliang](https://github.com/wgliang))

<code>string</code> takes the place of <code>P</code>.
And since <code>string</code> is identical to <code>string</code>,
this unification step succeeds as well.
Unification of the LHS and RHS of the equation is now finished.
Type inference succeeds because there is only one type equation,
no unification step failed, and the map is fully populated.
</p>

<p>
Unification uses a combination of <i>exact</i> and <i>loose</i>

performing the TLS Bootstrap properly. Earlier, `kubeadm join` only did the discovery part and exited successfully without checking that the kubelet actually started properly and performed the TLS bootstrap correctly. Now, as kubeadm runs some post-join steps (for example, annotating the Node API object with the CRISocket), `kubeadm join` is now waiting for the kubelet to perform the TLS Bootstrap, and then uses that credential to perform further actions. This also improves the UX, as `kubeadm` will exit...

Kubeadm: Expose the `kubeadm join` workflow as phases

- The `kubeadm join` command can now be used in phases. Similar to the work that was done for `kubeadm init` in 1.13, in 1.14 the `join` phases can be now executed step-by-step/selectively using the `kubeadm join phase` sub-command. This makes it possible to further customize the workflow of joining nodes to the cluster.
kubernetes/kubeadm: [#1204](https://github.com/kubernetes/kubeadm/issues/1204)

- Restored `NumPDBViolations` info of nodes, when `HTTPExtender ProcessPreemption`. This info will be used in subsequent filtering steps - `pickOneNodeForPreemption` ([#105853](https://github.com/kubernetes/kubernetes/pull/105853), [@caden2016](https://github.com/caden2016))

- Added enforcement of an upper cost bound for DRA evaluations of CEL. The API server and scheduler now enforce an upper bound on the cost and runtime steps required for evaluating a CEL expression. ([#128101](https://github.com/kubernetes/kubernetes/pull/128101), [@pohly](https://github.com/pohly)) [SIG API Machinery and Node]

The `wait.NewExponentialBackoffManager` and `wait.NewJitteringBackoffManager` functions have been marked as deprecated.  Callers should switch to using the `Backoff{...}.DelayWithReset(clock, resetInterval)` method and must set the `Steps` field when using `Factor`. As a short term change, callers may use the `Timer()` method on the `BackoffManager` until the backoff managers are deprecated and removed. Please see the godoc of the deprecated functions for examples of how to replace...

  To avoid breaking existing pods that use matchLabelKeys, the current scheduler behavior will be preserved until it is removed in v1.34. 
  Therefore, do not upgrade your scheduler directly from v1.32 to v1.34. 
  Instead, upgrade step-by-step (from v1.32 to v1.33, then to v1.34), 
  ensuring that any pods created at v1.32 with matchLabelKeys are either removed or already scheduled by the time you reach v1.34.
  

In v1.9 SIG AWS has improved stability of EBS support across the board. If a Volume is “stuck” in the attaching state to a node for too long a unschedulable taint will be applied to the node, so a Kubernetes admin can [take manual steps to correct the error](http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ebs-attaching-volume.html). Users are encouraged to ensure they are monitoring for the taint, and should consider automatically terminating instances in this state.

- Kube-controller-manager: the dynamic resource controller steps in when a pod got created such that the scheduler ignores it (i.e. spec.nodeName is set) and then takes care of triggering delayed resource claim allocation and/or reserving a claim for the pod. ([#118209](https://github.com/kubernetes/kubernetes/pull/118209),...