kubevirt: Ignore completed virt-launcher pods

When a live migration fails at completed virt-launcher pod with newer
creation timestamp is lingering in the system so the virt-launcher that
is really running the VM is hidden by it and routes are wrongly
reconstructed. This change take into account if the pod is completed to
ignore it marking it as stale.

Signed-off-by: Enrique Llorente <ellorent@redhat.com>

go-controller/pkg/kubevirt/pod.go

@@ -123,15 +123,18 @@ func EnsurePodAnnotationForVM(watchFactory *factory.WatchFactory, kube *kube.Kub
 	return podAnnotation, nil
 }
 
-// IsMigratedSourcePodStale return true if there are other pods related to
-// to it and any of them has newer creation timestamp.
+// IsMigratedSourcePodStale return false is the pod is live migratable and not completed
+// and is the running vm pod with newest creation timestamp
 func IsMigratedSourcePodStale(client *factory.WatchFactory, pod *corev1.Pod) (bool, error) {
+	if util.PodCompleted(pod) {
+		return true, nil
+	}
 	if !IsPodLiveMigratable(pod) {
-		return false, nil
+		return true, nil
 	}
 	vmPods, err := findVMRelatedPods(client, pod)
 	if err != nil {
-		return false, fmt.Errorf("failed finding related pods for pod %s/%s when checking live migration left overs: %v", pod.Namespace, pod.Name, err)
+		return true, fmt.Errorf("failed finding related pods for pod %s/%s when checking live migration left overs: %v", pod.Namespace, pod.Name, err)
 	}
 
 	for _, vmPod := range vmPods {