diff --git a/pkg/cluster/cluster.go b/pkg/cluster/cluster.go
index 75be4660a..5cc73c04d 100644
--- a/pkg/cluster/cluster.go
+++ b/pkg/cluster/cluster.go
@@ -9,7 +9,6 @@ import (
 	"fmt"
 	"reflect"
 	"regexp"
-	"strconv"
 	"sync"
 	"time"
 
@@ -680,24 +679,7 @@ func (c *Cluster) Update(oldSpec, newSpec *acidv1.Postgresql) error {
 	}()
 
 	if c.OpConfig.EnableUnusedPVCDeletion && oldSpec.Spec.NumberOfInstances > newSpec.Spec.NumberOfInstances {
-		c.logger.Debug("deleting pvc of shut down pods")
-
-		for i := oldSpec.Spec.NumberOfInstances - 1; i >= newSpec.Spec.NumberOfInstances; i-- {
-
-			// Scaling down to 0 replicas is not cluster deletion so keep the last pvc.
-			// Operator will remove it only when explicit "kubectl pg delete" is issued
-			if i == 0 {
-				c.logger.Info("cluster scaled down to 0 pods; skipping deletion of the last pvc")
-				break
-			}
-
-			podIndex := strconv.Itoa(int(i))
-			pvcName := "pgdata-" + c.Name + "-" + podIndex
-			if err := c.KubeClient.PersistentVolumeClaims(c.Namespace).Delete(context.TODO(), pvcName, c.deleteOptions); err != nil {
-				c.logger.Warningf("could not delete PersistentVolumeClaim: %v", err)
-				// failing to delete pvc does not fail the update; Sync() may also delete unused PVCs later
-			}
-		}
+		c.deleteUnusedPersistentVolumeClaims()
 	}
 
 	// pod disruption budget
diff --git a/pkg/cluster/sync.go b/pkg/cluster/sync.go
index 271c03dba..b68abfb16 100644
--- a/pkg/cluster/sync.go
+++ b/pkg/cluster/sync.go
@@ -3,7 +3,6 @@ package cluster
 import (
 	"context"
 	"fmt"
-	"strconv"
 
 	batchv1beta1 "k8s.io/api/batch/v1beta1"
 	v1 "k8s.io/api/core/v1"
@@ -114,20 +113,7 @@ func (c *Cluster) Sync(newSpec *acidv1.Postgresql) error {
 	// remove unused PVCs in case deleting them during scale down failed; see Update()
 	// the last pvc stays until the cluster is explicitly deleted as opposed to being scaled down to 0 pods
 	if c.OpConfig.EnableUnusedPVCDeletion && c.getNumberOfInstances(&c.Spec) > 0 {
-
-		// XXX that also deletes PVC of pods shut down before this change is deployed
-		for i := c.getNumberOfInstances(&c.Spec); ; i++ {
-			podIndex := strconv.Itoa(int(i))
-			pvcName := "pgdata-" + c.Name + "-" + podIndex
-			if err := c.KubeClient.PersistentVolumeClaims(c.Namespace).Delete(context.TODO(), pvcName, c.deleteOptions); err != nil {
-				if k8sutil.ResourceNotFound(err) {
-					// no more pvcs to delete
-					break
-				}
-				c.logger.Warningf("could not delete PersistentVolumeClaim: %v", err)
-				// next Sync() or Update() will retry
-			}
-		}
+		c.deleteUnusedPersistentVolumeClaims()
 	}
 
 	// sync connection pooler
diff --git a/pkg/cluster/volumes.go b/pkg/cluster/volumes.go
index a5bfe6c2d..084bd2484 100644
--- a/pkg/cluster/volumes.go
+++ b/pkg/cluster/volumes.go
@@ -15,6 +15,7 @@ import (
 	"github.com/zalando/postgres-operator/pkg/util"
 	"github.com/zalando/postgres-operator/pkg/util/constants"
 	"github.com/zalando/postgres-operator/pkg/util/filesystems"
+	"github.com/zalando/postgres-operator/pkg/util/k8sutil"
 	"github.com/zalando/postgres-operator/pkg/util/volumes"
 )
 
@@ -52,6 +53,33 @@ func (c *Cluster) deletePersistentVolumeClaims() error {
 	return nil
 }
 
+func (c *Cluster) deleteUnusedPersistentVolumeClaims() {
+
+	c.logger.Debug("deleting pvc of shut down pods")
+
+	// Scaling down to 0 replicas is not cluster deletion so keep the last pvc.
+	// Operator will remove it only when explicit "kubectl pg delete" is issued
+	if c.getNumberOfInstances(&c.Spec) == 0 {
+		c.logger.Info("cluster scaled down to 0 pods; skipping deletion of the last pvc")
+		return
+	}
+
+	// XXX that also deletes PVC of pods shut down before this change is deployed
+	for i := c.getNumberOfInstances(&c.Spec); ; i++ {
+		podIndex := strconv.Itoa(int(i))
+		pvcName := "pgdata-" + c.Name + "-" + podIndex
+		if err := c.KubeClient.PersistentVolumeClaims(c.Namespace).Delete(context.TODO(), pvcName, c.deleteOptions); err != nil {
+			if k8sutil.ResourceNotFound(err) {
+				// no more pvcs to delete
+				break
+			}
+			c.logger.Warningf("could not delete PersistentVolumeClaim: %v", err)
+			// next Sync() or Update() will retry
+		}
+	}
+
+}
+
 func (c *Cluster) listPersistentVolumes() ([]*v1.PersistentVolume, error) {
 	result := make([]*v1.PersistentVolume, 0)