`autoSizingReserved: true` feature reserves excessively low (i.e. 90 millicores vs. 500 millicores default where autosizing is disabled) on control plane nodes, leading to inadequate systemReserved allocation and disrupting control plane operations

OCP 4.16

Always

1. enable `autoSizingReserved: true` feature in KubeletConfig
2. check CPU reservation on nodes /etc/node-sizing.env after KubeletConfig applied

https://docs.openshift.com/container-platform/4.10/nodes/nodes/nodes-nodes-resources-configuring.html#nodes-nodes-resources-configuring-auto_nodes-nodes-resources-configuring 

$ awk 'NR == 1 || $3 ~ "2805"' sos_commands/process/ps_alxwww
F   UID     PID    PPID PRI  NI    VSZ   RSS WCHAN  STAT TTY        TIME COMMAND
4     0    2805       1  20   0 5164428 339568 -    Rsl  ?        125:45 /usr/bin/kubelet --config=/etc/kubernetes/kubelet.conf --bootstrap-kubeconfig=/etc/kubernetes/kubeconfig --kubeconfig=/var/lib/kubelet/kubeconfig --container-runtime-endpoint=/var/run/crio/crio.sock --runtime-cgroups=/system.slice/crio.service --node-labels=node-role.kubernetes.io/control-plane,node-role.kubernetes.io/master,node.openshift.io/os_id=rhcos --node-ip=0.0.0.0 --minimum-container-ttl-duration=6m0s --cloud-provider=external --volume-plugin-dir=/etc/kubernetes/kubelet-plugins/volume/exec --image-credential-provider-bin-dir=/usr/libexec/kubelet-image-credential-provider-plugins --image-credential-provider-config=/etc/kubernetes/credential-providers/acr-credential-provider.yaml --hostname-override= --provider-id= --register-with-taints=node-role.kubernetes.io/master=:NoSchedule --pod-infra-container-image=quay.io/openshift-release-dev/ocp-v4.0-art-dev@sha256:e6485f12a32f3b065c6e9ada0f0cd0a766def9458e88ab423cac16a6ba6e9cc7 --system-reserved=cpu=0.09,memory=5Gi,ephemeral-storage=1Gi --v=2

`autoSizingReserved` should not reserve lesser CPU than would be reserved by default without any KubeletConfig in place. Allocations calculated by `autoSizingReserved` feature should add to the base CPU reservation of 500m

In OpenShift Container Platform 4.10 (and newer), half of a CPU core (500 millicore) is now reserved by the system by default to the system slice. However, when `autoSizingReserved: true` is set, this is significantly reduced to 90 millicores... 

https://docs.openshift.com/container-platform/4.10/scalability_and_performance/recommended-host-practices.html#master-node-sizing_recommended-host-practices