- Clone `kubernetes/kubernetes` in `update-vendor.sh` shallowly, instead of fetching all revisions
  - Speed up binpacking by reducing the number of PreFilter calls (call once per pod instead of #pods*#nodes times)
  - Speed up finding unneeded nodes by 5x+ in very large clusters by reducing the number of PreFilter calls
  - Expose `--max-nodes-total` as a metric
  - Errors in `IncreaseSize` changed from type `apiError` to `cloudProviderError`

* Add generic NoSchedule toleration to fluentd in gcp config. ([#48182](https://github.com/kubernetes/kubernetes/pull/48182), [@gmarek](https://github.com/gmarek))
* RBAC role and role-binding reconciliation now ensures namespaces exist when reconciling on startup. ([#48480](https://github.com/kubernetes/kubernetes/pull/48480), [@liggitt](https://github.com/liggitt))

labels.facet_filetype_ods=filetype:ods\t\
labels.facet_filetype_odp=filetype:odp\t\
labels.facet_filetype_pdf=filetype:pdf\t\
labels.facet_filetype_txt=filetype:txt\t\
labels.facet_filetype_others=filetype:others\n\

# ranking

# Window size for rank fusion.
rank.fusion.window_size=200
# Rank constant for rank fusion.
rank.fusion.rank_constant=20
# Number of threads for rank fusion.
rank.fusion.threads=-1

  * Marking a StorageClass as default makes it so that even a PersistentVolumeClaim without a StorageClass specified will trigger dynamic provisioning (instead of binding to an existing pool of PVs).
  * If you depend on the old behavior of volumes binding to existing pool of PersistentVolume objects then modify the StorageClass object and set `storageclass.beta.kubernetes.io/is-default-class` to `false`.

* Fixed binding of block PersistentVolumes / PersistentVolumeClaims when BlockVolume feature is off. ([#84049](https://github.com/kubernetes/kubernetes/pull/84049), [@jsafrane](https://github.com/jsafrane))

- Fixed a regression preventing garbage collection of RBAC role and binding objects ([#90534](https://github.com/kubernetes/kubernetes/pull/90534), [@apelisse](https://github.com/apelisse)) [SIG Auth]

* Fix kube-proxy PodSecurityPolicy binding on GCE & GKE. This was only an issue when running kube-proxy as a DaemonSet, with PodSecurityPolicy enabled. ([#72761](https://github.com/kubernetes/kubernetes/pull/72761), [@tallclair](https://github.com/tallclair))

* Initialize podsWithAffinity to avoid scheduler panic ([#33967](https://github.com/kubernetes/kubernetes/pull/33967), [@xiang90](https://github.com/xiang90))
* Fix base image pinning during upgrades via cluster/gce/upgrade.sh ([#33147](https://github.com/kubernetes/kubernetes/pull/33147), [@vishh](https://github.com/vishh))

  The new feature gate activates a framework for fine-grained filtering of events related to scheduler plugins.
  In this release, no default scheduling plugins make use of the hinting framework, so you should not expect any behavior changes. ([#119328](https://github.com/kubernetes/kubernetes/pull/119328), [@sanposhiho](https://github.com/sanposhiho)) [SIG Scheduling]

- Added kubelet metrics to track the memory manager allocation and pinning. ([#121778](https://github.com/kubernetes/kubernetes/pull/121778), [@Tal-or](https://github.com/Tal-or))