I recently noticed that cluster-autoscaler pod in the hosted control plane namespace is going continuous restarts. Upon observing the issue, found out liveness and readiness probe failing on this pod. 

Also, checking further the logs of this pod, points to rbac missing for the cluster-autoscaler pod in this case. Please see the logs trace for reference. 

E0215 14:52:59.936182 1 reflector.go:140] k8s.io/client-go/dynamic/dynamicinformer/informer.go:91: Failed to watch *unstructured.Unstructured: failed to list *unstructured.Unstructured: agentmachinetemplates.capi-provider.agent-install.openshift.io is forbidden: User "system:serviceaccount:clusters-hcp01:cluster-autoscaler" cannot list resource "agentmachinetemplates" in API group "capi-provider.agent-install.openshift.io" in the namespace "clusters-hcp01"