csi: update csi holder daemonset template

Currently the holder daemonset is never updated
which will leaves the images the daemonset also
not updated. we should update the daemonset
template but not restart the csi holder pods
which causes the CSI volume access problem,
set the updateStrategy to OnDelete (already set in
yaml files) which allow us to update the holder
daemonset but not restart/update the pods, when a
pod is deleted or node is rebooted
the new changes will take effect.

Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
(cherry picked from commit 1ba3aa4d19)

pkg/operator/ceph/csi/spec.go

@@ -870,7 +870,11 @@ func (r *ReconcileCSI) configureHolder(driver driverDetails, c ClusterDetail, tp
 	_, err = r.context.Clientset.AppsV1().DaemonSets(r.opConfig.OperatorNamespace).Create(r.opManagerContext, cephPluginHolder, metav1.CreateOptions{})
 	if err != nil {
 		if kerrors.IsAlreadyExists(err) {
-			logger.Debugf("holder %q already exists for cluster %q, it should never be updated", cephPluginHolder.Name, c.cluster.Namespace)
+			_, err = r.context.Clientset.AppsV1().DaemonSets(r.opConfig.OperatorNamespace).Update(r.opManagerContext, cephPluginHolder, metav1.UpdateOptions{})
+			if err != nil {
+				return errors.Wrapf(err, "failed to update ceph plugin holder daemonset %q", cephPluginHolder.Name)
+			}
+			logger.Debugf("holder %q already exists for cluster %q, updating it, restart holder pods to take effect of update", cephPluginHolder.Name, c.cluster.Namespace)
 		} else {
 			return errors.Wrapf(err, "failed to start ceph plugin holder daemonset %q", cephPluginHolder.Name)
 		}