When running a vdu test app workload on an SNO with DU profile running non-rt kernel the load average increases to ~130. While the node is under load, some of the times the kube api cannot recover following a rollout.

4.12.0-rc.0 with 4.18.0-372.36.1.el8_6.x86_64 kernel

Consistently

1. Deploy and configure SNO with DU profile with 4.18.0-372.36.1.el8_6.x86_64 kernel

2. Deploy test app

3. Force a kube api server rollout:
oc patch kubeapiserver cluster -p='{"spec": {"forceRedeploymentReason": "recovery-'"$( date --rfc-3339=ns )"'"}}' --type=merge 

4. Wait for kube-apiserver to achieve a new revision

Some of the times the kube-apiserver doesn't recover and remains unreachable.

kube-apiserver always recovers