#### Network

- The deprecated flag `--conntrack-max` has been removed from kube-proxy. Users of this flag should switch to `--conntrack-min` and `--conntrack-max-per-core` instead. ([#78399](https://github.com/kubernetes/kubernetes/pull/78399), [@rikatz](https://github.com/rikatz))

* Fix GPU resource validation ([#28743](https://github.com/kubernetes/kubernetes/pull/28743), [@therc](https://github.com/therc))
* Scale kube-proxy conntrack limits by cores (new default behavior) ([#28876](https://github.com/kubernetes/kubernetes/pull/28876), [@thockin](https://github.com/thockin))

### Other notable changes

* Fix unintended change of Service.spec.ports[].nodePort during kubectl apply ([#24180](https://github.com/kubernetes/kubernetes/pull/24180), [@AdoHe](https://github.com/AdoHe))
* Flush conntrack state for removed/changed UDP Services ([#22573](https://github.com/kubernetes/kubernetes/pull/22573), [@freehan](https://github.com/freehan))

  fix: ensure InstanceShutdownByProviderID return false for creating Azure VMs ([#104448](https://github.com/kubernetes/kubernetes/pull/104448), [@feiskyer](https://github.com/feiskyer)) [SIG Cloud Provider]
- Kube-proxy: delete stale conntrack UDP entries for loadbalancer ingress IP. ([#104152](https://github.com/kubernetes/kubernetes/pull/104152), [@aojea](https://github.com/aojea)) [SIG Network]

* Reduce CSI log and event spam. ([#71581](https://github.com/kubernetes/kubernetes/pull/71581), [@saad-ali](https://github.com/saad-ali))
* Add conntrack as a dependency of kubelet and kubeadm when building rpms and debs. Both require conntrack to handle cleanup of connections. ([#71540](https://github.com/kubernetes/kubernetes/pull/71540), [@mauilion](https://github.com/mauilion))

- github.com/moby/term: [df9cb8a → 9d4ed18](https://github.com/moby/term/compare/df9cb8a...9d4ed18)
- github.com/mwitkow/go-conntrack: [cc309e4 → 2f06839](https://github.com/mwitkow/go-conntrack/compare/cc309e4...2f06839)
- github.com/onsi/ginkgo: [v1.11.0 → v1.14.0](https://github.com/onsi/ginkgo/compare/v1.11.0...v1.14.0)

* Fix race condition in setting node statusUpdateNeeded flag  ([#32807](https://github.com/kubernetes/kubernetes/pull/32807), [@jingxu97](https://github.com/jingxu97))
* kube-proxy: Add a lower-bound for conntrack (128k default) ([#33051](https://github.com/kubernetes/kubernetes/pull/33051), [@thockin](https://github.com/thockin))

- Kubelet TCP and HTTP probes are now more effective using networking resources:
  conntrack entries, sockets. This is achieved by reducing the `TIME-WAIT` state
  of the connection to 1 second, instead of the defaults 60 seconds. This allows
  kubelet to free the socket, and free conntrack entry and ephemeral port associated. ([#115143](https://github.com/kubernetes/kubernetes/pull/115143), [@aojea](https://github.com/aojea))

- Kube-proxy: delete stale conntrack UDP entries for loadbalancer ingress IP. ([#104151](https://github.com/kubernetes/kubernetes/pull/104151), [@aojea](https://github.com/aojea)) [SIG Network]

## Changes by Kind

### Feature

- Kubelet TCP and HTTP probes are more effective using networking resources: conntrack entries, sockets, ...