* Scale up forbidden window was protecting HPA against making decision to scale up based on metrics gathered during pod initialisation (which may be invalid, for example pod may be using a lot of CPU despite not doing any "actual" work).
    * To avoid that negative effect only use per pod metrics from pods that are:

- Fixed a potential memory leak in QueueingHint (alpha feature). ([#127016](https://github.com/kubernetes/kubernetes/pull/127016), [@sanposhiho](https://github.com/sanposhiho))
- Fixed a race condition in the kube-proxy initialization that could cause UDP traffic to service VIP. ([#126532](https://github.com/kubernetes/kubernetes/pull/126532), [@wedaly](https://github.com/wedaly))

 - Base-images: Update to kube-cross:v1.13.9-5 ([#90964](https://github.com/kubernetes/kubernetes/pull/90964), [@justaugustus](https://github.com/justaugustus)) [SIG Release and Testing]
 - CSINode initialization does not crash kubelet on startup when APIServer is not reachable or kubelet has not the right credentials yet. ([#89589](https://github.com/kubernetes/kubernetes/pull/89589), [@jsafrane](https://github.com/jsafrane)) [SIG Storage]

* Fixed a bug in the API server watch cache, which could cause a missing watch event immediately after cache initialization. ([#49992](https://github.com/kubernetes/kubernetes/pull/49992), [@liggitt](https://github.com/liggitt))



# v1.6.8

- Fix race condition in kube-proxy initialization that could blackhole UDP traffic to service VIP. ([#126688](https://github.com/kubernetes/kubernetes/pull/126688), [@wedaly](https://github.com/wedaly)) [SIG Network]

* Prevent kubelet from getting wedged if initialization of modules returns an error. ([#59020](https://github.com/kubernetes/kubernetes/pull/59020), [@brendandburns](https://github.com/brendandburns))

* Cluster Autoscaler 1.0.4 ([#59271](https://github.com/kubernetes/kubernetes/pull/59271), [@mwielgus](https://github.com/mwielgus))
* Prevent kubelet from getting wedged if initialization of modules returns an error. ([#59020](https://github.com/kubernetes/kubernetes/pull/59020), [@brendandburns](https://github.com/brendandburns))

* Fix concurrent map access panic ([#65327](https://github.com/kubernetes/kubernetes/pull/65327), [@dashpole](https://github.com/dashpole))
    * Don't watch .mount cgroups to reduce number of inotify watches
    * Fix NVML initialization race condition
    * Fix brtfs disk metrics when using a subdirectory of a subvolume

* Fix concurrent map access panic ([#65328](https://github.com/kubernetes/kubernetes/pull/65328), [@dashpole](https://github.com/dashpole))
    * Don't watch .mount cgroups to reduce number of inotify watches
    * Fix NVML initialization race condition
    * Fix brtfs disk metrics when using a subdirectory of a subvolume
* Pass cluster_location argument to Heapster ([#65176](https://github.com/kubernetes/kubernetes/pull/65176), [@kawych](https://github.com/kawych))

- Kubeadm: graduated the kubeadm specific feature gate `WaitForAllControlPlaneComponents` to GA. The feature gate is was locked to always be enabled and on node initialization kubeadm performed a health check for all control plane components and not only the `kube-apiserver`. ([#132594](https://github.com/kubernetes/kubernetes/pull/132594), [@neolit123](https://github.com/neolit123))