* Update Azure load balancer to prevent orphaned public IP addresses ([#82890](https://github.com/kubernetes/kubernetes/pull/82890), [@chewong](https://github.com/chewong))
* Update Cluster Autoscaler to 1.15.2 (https://github.com/kubernetes/autoscaler/releases/tag/cluster-autoscaler-1.15.2) ([#82831](https://github.com/kubernetes/kubernetes/pull/82831), [@krzysztof-jastrzebski](https://github.com/krzysztof-jastrzebski))

* Update Cluster Autoscaler to 1.14.2 ([#77064](https://github.com/kubernetes/kubernetes/pull/77064), [@losipiuk](https://github.com/losipiuk))
    * https://github.com/kubernetes/autoscaler/releases/tag/cluster-autoscaler-1.14.2 
    * https://github.com/kubernetes/autoscaler/releases/tag/cluster-autoscaler-1.14.1

* Update Cluster Autoscaler version to 1.16.2 (CA release docs: https://github.com/kubernetes/autoscaler/releases/tag/cluster-autoscaler-1.16.2) ([#84038](https://github.com/kubernetes/kubernetes/pull/84038), [@losipiuk](https://github.com/losipiuk))

* On GCP platforms, e2e testing now logs which OS images the cluster was found to have. ([#48310](https://github.com/kubernetes/kubernetes/pull/48310), [@abgworrall](https://github.com/abgworrall))

specified it was set to 80 by default while in autoscaling/v1 HPA object
without TargetCPUUtilizationPercentage specified is a valid object. Pod
autoscaler controller will apply a default scaling policy in this case which is
equivalent to the previous one but may change in the future.

- If `PreBindPreFlight` returned `Skip`, the scheduler didn't run the plugin at `PreBind`.
  If any `PreBindPreFlight` returned `Success`, the scheduler put NominatedNodeName to the pod 
  so that other components (such as the cluster autoscaler) could notice the pod was going to be bound to the node. ([#133021](https://github.com/kubernetes/kubernetes/pull/133021), [@sanposhiho](https://github.com/sanposhiho)) [SIG Scheduling and Testing]

- [KCCM - service controller]: enabled connection draining for terminating pods upon node downscale by the cluster autoscaler. This is done by not reacting to the taint used by the cluster autoscaler to indicate that the node is going away soon, thus keeping the node referenced by the load balancer until the VM has been completely deleted. ([#115204](https://github.com/kubernetes/kubernetes/pull/115204),...

- The scheduler now clears the `nominatedNodeName` field for Pods upon scheduling or binding failure. External components, such as Cluster Autoscaler and Karpenter, should not overwrite this field. ([#135007](https://github.com/kubernetes/kubernetes/pull/135007), [@ania-borowiec](https://github.com/ania-borowiec)) [SIG Scheduling and Testing]

- Updated Cluster Autosaler to version `1.22.0`. Release notes: https://github.com/kubernetes/autoscaler/releases/tag/cluster-autoscaler-1.22.0. ([#104293](https://github.com/kubernetes/kubernetes/pull/104293), [@x13n](https://github.com/x13n))

* Incompatible change of kube-up.sh:  ([#25734](https://github.com/kubernetes/kubernetes/pull/25734), [@jszczepkowski](https://github.com/jszczepkowski))
    * When turning on cluster autoscaler by setting KUBE_ENABLE_NODE_AUTOSCALER=true,KUBE_AUTOSCALER_MIN_NODES and KUBE_AUTOSCALER_MAX_NODES need to be set.