openshift: add PodOverhead to RuntimeClass definition

Upstream kata-deploy has the overhead of a Kata Pod
set as

overhead:
    podFixed:
        memory: "160Mi"
        cpu: "250m"

(https://github.com/kata-containers/packaging/blob/master/kata-deploy/k8s-1.18/kata-runtimeClasses.yaml#L7)

So far we had no Pod Overhead set at all. Let's use the same values
as upstream uses to take the Pod resource overhead into account when the
'kata' runtime class is used.

Signed-off-by: Jens Freimann <jfreimann@redhat.com>

controllers/openshift_controller.go

@@ -33,6 +33,7 @@ import (
 	corev1 "k8s.io/api/core/v1"
 	nodeapi "k8s.io/api/node/v1beta1"
 	"k8s.io/apimachinery/pkg/api/errors"
+	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/types"
@@ -447,6 +448,14 @@ func (r *KataConfigOpenShiftReconciler) setRuntimeClass() (ctrl.Result, error) {
 				Name: runtimeClassName,
 			},
 			Handler: runtimeClassName,
+			// Use same values for Pod Overhead as upstream kata-deploy using, see
+			// https://github.com/kata-containers/packaging/blob/f17450317563b6e4d6b1a71f0559360b37783e19/kata-deploy/k8s-1.18/kata-runtimeClasses.yaml#L7
+			Overhead: &nodeapi.Overhead{
+				PodFixed: corev1.ResourceList{
+					corev1.ResourceCPU:    resource.MustParse("250m"),
+					corev1.ResourceMemory: resource.MustParse("160Mi"),
+				},
+			},
 		}
 
 		if r.kataConfig.Spec.KataConfigPoolSelector != nil {