Use PodManagementPolicy: appsv1.ParallelPodManagement for cinder api statefulset

With the default PodManagementPolicy: OrderedReadyPodManagement
the statefulset controller will only progress pods when the
previous/current pod is ready or terminated.

When service configuration changes while the pod is starting
and the new configuration requires e.g. additional volume mounts
the initial pod will never reach ready and therefore an update
won't happen.

With ParallelPodManagement the statefulset controller will
not wait for pods to be ready or complete termination.

Author: stuggi

pkg/cinderapi/statefuleset.go

@@ -116,7 +116,8 @@ func StatefulSet(
 			Selector: &metav1.LabelSelector{
 				MatchLabels: labels,
 			},
-			Replicas: instance.Spec.Replicas,
+			PodManagementPolicy: appsv1.ParallelPodManagement,
+			Replicas:            instance.Spec.Replicas,
 			Template: corev1.PodTemplateSpec{
 				ObjectMeta: metav1.ObjectMeta{
 					Annotations: annotations,

test/kuttl/common/assert_sample_deployment.yaml

@@ -31,7 +31,7 @@ metadata:
     kind: CinderAPI
     name: cinder-api
 spec:
-  podManagementPolicy: OrderedReady
+  podManagementPolicy: Parallel
   replicas: 1
   revisionHistoryLimit: 10
   selector: