Packages changed:
  kubernetes (1.21.3 -> 1.22.1)
  kubernetes1.21 (1.21.3 -> 1.21.4)

=== Details ===

==== kubernetes ====
Version update (1.21.3 -> 1.22.1)
Subpackages: kubernetes-client kubernetes-kubeadm kubernetes-kubelet

- Bump coredns version to 1.8.4 and etcd to 3.5.0
- Bump kubernetes-* to 1.22.1 and kubernetes-*-minus1 to 1.21.4

==== kubernetes1.21 ====
Version update (1.21.3 -> 1.21.4)

- Update to version 1.21.4:
  * Avoid spurious calls to update/delete validation
  * Update to go1.16.7
  * Fix metrics reporting for the deprecated watch path
  * Update configure-helper.sh
  * Update configure-helper.sh
  * Update configure-helper.sh
  * Fix disruptive subPath test failures
  * Fix: ignore not a VMSS error for VMAS nodes in reconcileBackendPools
  * feat: Provide IPv6 support for internal load balancer
  * Update to using apiserver-network-proxy v1.22
  * Make CSR cleaner tolerate objects with invalid status.certificate
  * disable aufs module
  * Update pd csi driver images to use v1 images (CSINode, CSIDriver, etc)
  * storage e2e: patch in RBAC rules for secrets
  * storage e2e: downgrade hostpath driver
  * storage e2e: disable health check containers
  * storage e2e: automate hostpath YAML updates, hostpath v1.6.2
  * update comments and owners file for pkg/util/removeall
  * kubelet: do not call RemoveAll on volumes directory for orphaned pods
  * Fix panic in master upgrade tests
  * Update CHANGELOG/CHANGELOG-1.21.md for v1.21.3
  * Updated to use konnectivity client v0.0.21, and implemented placeholder context
  * Fix the code is leaking the defaulting between unrelated pod instances
  * Simplify use of the fake dynamic client
  * fix: return empty VMAS name if using standalone VM
  * wait for endpoints to be available
  * Remove extra zone test
  * Fix frameworkImpl.extenders being not set
  * fix: Refresh VM cache when node is not found
  * fix: cleanup outdated routes
  * strip prefix 'v' in coredns version if exists
  * Fix closing of decorated watcher channel on timeout
  * fix: delete non existing disk issue
  * let objects without metadata pass through the managedFields admission controller