fix: prevent unnecessary rolling restarts during node draining

opensearch-operator/pkg/helpers/helpers.go

@@ -482,6 +482,11 @@ func ReplicaHostName(currentSts appsv1.StatefulSet, repNum int32) string {
 }
 
 func WorkingPodForRollingRestart(k8sClient k8s.K8sClient, sts *appsv1.StatefulSet) (string, error) {
+	// If revisions match (no spec change), indicate no work without requeue-triggering error
+	if sts.Status.UpdateRevision != "" && sts.Status.CurrentRevision == sts.Status.UpdateRevision {
+		return "", nil
+	}
+
 	// If there are potentially mixed revisions we need to check each pod
 	podWithOlderRevision, err := GetPodWithOlderRevision(k8sClient, sts)
 	if err != nil {
@@ -490,7 +495,9 @@ func WorkingPodForRollingRestart(k8sClient k8s.K8sClient, sts *appsv1.StatefulSe
 	if podWithOlderRevision != nil {
 		return podWithOlderRevision.Name, nil
 	}
-	return "", errors.New("unable to calculate the working pod for rolling restart")
+
+	// No pod requires restart at the moment; return empty name with nil error
+	return "", nil
 }
 
 // DeleteStuckPodWithOlderRevision deletes the crashed pod only if there is any update in StatefulSet.

opensearch-operator/pkg/reconcilers/rollingRestart.go

@@ -74,6 +74,15 @@ func (r *RollingRestartReconciler) Reconcile() (ctrl.Result, error) {
 		if err != nil {
 			return ctrl.Result{}, err
 		}
+
+		// Skip rolling restart if revisions are the same (no actual spec changes)
+		// This prevents unnecessary rolling restarts during node draining/pod recreation
+		if sts.Status.UpdateRevision != "" && sts.Status.CurrentRevision == sts.Status.UpdateRevision {
+			// All pods are up to date, no rolling restart needed
+			lg.V(1).Info(fmt.Sprintf("StatefulSet %s has matching current and update revisions, skipping rolling restart", sts.Name))
+			continue
+		}
+
 		if sts.Status.UpdateRevision != "" &&
 			sts.Status.UpdatedReplicas != ptr.Deref(sts.Spec.Replicas, int32(1)) {
 			pendingUpdate = true
@@ -147,6 +156,14 @@ func (r *RollingRestartReconciler) Reconcile() (ctrl.Result, error) {
 		if err != nil {
 			return ctrl.Result{}, err
 		}
+
+		// Skip rolling restart if revisions are the same (no actual spec changes)
+		// This prevents unnecessary rolling restarts during node draining/pod recreation
+		if sts.Status.UpdateRevision != "" && sts.Status.CurrentRevision == sts.Status.UpdateRevision {
+			lg.V(1).Info(fmt.Sprintf("StatefulSet %s has matching current and update revisions, skipping rolling restart", sts.Name))
+			continue
+		}
+
 		if sts.Status.UpdateRevision != "" &&
 			sts.Status.UpdatedReplicas != ptr.Deref(sts.Spec.Replicas, 1) {
 			// Only restart pods if not all pods are updated and the sts is healthy with no pods terminating
@@ -190,6 +207,10 @@ func (r *RollingRestartReconciler) restartStatefulSetPod(sts *appsv1.StatefulSet
 	if err != nil {
 		return ctrl.Result{}, err
 	}
+	if workingPod == "" {
+		// Nothing to restart right now; do not requeue due to this condition
+		return ctrl.Result{}, nil
+	}
 
 	lg.Info(fmt.Sprintf("Preparing to restart pod %s", workingPod))
 	ready, err = services.PreparePodForDelete(r.osClient, lg, workingPod, r.instance.Spec.General.DrainDataNodes, dataCount)

opensearch-operator/pkg/reconcilers/upgrade.go

@@ -348,6 +348,11 @@ func (r *UpgradeReconciler) doNodePoolUpgrade(pool opsterv1.NodePool) error {
 		r.setComponentConditions(conditions, pool.Component)
 		return err
 	}
+	if workingPod == "" {
+		// No pod to restart right now; this is a non-error state
+		r.setComponentConditions(conditions, pool.Component)
+		return nil
+	}
 
 	ready, err = services.PreparePodForDelete(r.osClient, r.logger, workingPod, r.instance.Spec.General.DrainDataNodes, dataCount)
 	if err != nil {