YDBOPS-7162 Immutable statefulSet selectorLabels with label "ydb.tech/statefulset-name"

deploy/ydb-operator/Chart.yaml

@@ -15,10 +15,10 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: 0.5.15
+version: 0.5.16
 
 # This is the version number of the application being deployed. This version number should be
 # incremented each time you make changes to the application. Versions are not expected to
 # follow Semantic Versioning. They should reflect the version the application is using.
 # It is recommended to use it with quotes.
-appVersion: "0.5.15"
+appVersion: "0.5.16"

internal/controllers/database/sync.go

@@ -322,42 +322,32 @@ func (r *Reconciler) waitForStatefulSetToScale(
 		r.Recorder.Event(
 			database,
 			corev1.EventTypeWarning,
-			"ProvisioningFailed",
-			fmt.Sprintf("Failed to get StatefulSets: %s", err),
+			"ControllerError",
+			fmt.Sprintf("Failed to get StatefulSet: %s", err),
 		)
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 	}
 
-	podLabels := labels.Common(database.Name, make(map[string]string))
-	podLabels.Merge(map[string]string{
-		labels.ComponentKey: labels.DynamicComponent,
-	})
-
-	matchingLabels := client.MatchingLabels{}
-	for k, v := range podLabels {
-		matchingLabels[k] = v
-	}
-
 	podList := &corev1.PodList{}
 	opts := []client.ListOption{
 		client.InNamespace(database.Namespace),
-		matchingLabels,
+		client.MatchingLabels{labels.StatefulsetComponent: database.Name},
 	}
 
 	err = r.List(ctx, podList, opts...)
 	if err != nil {
 		r.Recorder.Event(
 			database,
 			corev1.EventTypeWarning,
-			"ProvisioningFailed",
-			fmt.Sprintf("Failed to list cluster pods: %s", err),
+			"ControllerError",
+			fmt.Sprintf("Failed to list StatefulSet pods: %s", err),
 		)
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 	}
 
 	runningPods := 0
 	for _, e := range podList.Items {
-		if e.Status.Phase == "Running" {
+		if resources.PodIsReady(e) {
 			runningPods++
 		}
 	}
@@ -367,13 +357,13 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			database,
 			corev1.EventTypeNormal,
 			string(DatabaseProvisioning),
-			fmt.Sprintf("Waiting for number of running dynamic pods to match expected: %d != %d", runningPods, database.Spec.Nodes),
+			fmt.Sprintf("Waiting for number of running pods to match expected: %d != %d", runningPods, database.Spec.Nodes),
 		)
 		meta.SetStatusCondition(&database.Status.Conditions, metav1.Condition{
 			Type:    DatabaseProvisionedCondition,
 			Status:  metav1.ConditionFalse,
 			Reason:  ReasonInProgress,
-			Message: fmt.Sprintf("Number of running dynamic pods does not match expected: %d != %d", runningPods, database.Spec.Nodes),
+			Message: fmt.Sprintf("Number of running pods does not match expected: %d != %d", runningPods, database.Spec.Nodes),
 		})
 		return r.updateStatus(ctx, database, DefaultRequeueDelay)
 	}
@@ -383,7 +373,7 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			Type:    DatabaseProvisionedCondition,
 			Status:  metav1.ConditionTrue,
 			Reason:  ReasonCompleted,
-			Message: "Successfully scaled to desired number of nodes",
+			Message: fmt.Sprintf("Successfully scaled to desired number of nodes: %d", database.Spec.Nodes),
 		})
 		return r.updateStatus(ctx, database, StatusUpdateRequeueDelay)
 	}

internal/controllers/databasenodeset/sync.go

@@ -19,6 +19,7 @@ import (
 
 	"github.com/ydb-platform/ydb-kubernetes-operator/api/v1alpha1"
 	. "github.com/ydb-platform/ydb-kubernetes-operator/internal/controllers/constants" //nolint:revive,stylecheck
+	"github.com/ydb-platform/ydb-kubernetes-operator/internal/labels"
 	"github.com/ydb-platform/ydb-kubernetes-operator/internal/resources"
 )
 
@@ -213,15 +214,15 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			r.Recorder.Event(
 				databaseNodeSet,
 				corev1.EventTypeWarning,
-				"Syncing",
-				fmt.Sprintf("Failed to found StatefulSet: %s", err),
+				"ProvisioningFailed",
+				fmt.Sprintf("StatefulSet with name %s was not found: %s", databaseNodeSet.Name, err),
 			)
 			return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, nil
 		}
 		r.Recorder.Event(
 			databaseNodeSet,
 			corev1.EventTypeWarning,
-			"Syncing",
+			"ControllerError",
 			fmt.Sprintf("Failed to get StatefulSets: %s", err),
 		)
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
@@ -235,21 +236,22 @@ func (r *Reconciler) waitForStatefulSetToScale(
 	podList := &corev1.PodList{}
 	opts := []client.ListOption{
 		client.InNamespace(databaseNodeSet.Namespace),
-		matchingLabels,
+		client.MatchingLabels{labels.StatefulsetComponent: databaseNodeSet.Name},
 	}
-	if err = r.List(ctx, podList, opts...); err != nil {
+	err = r.List(ctx, podList, opts...)
+	if err != nil {
 		r.Recorder.Event(
 			databaseNodeSet,
 			corev1.EventTypeWarning,
-			"Syncing",
-			fmt.Sprintf("Failed to list databaseNodeSet pods: %s", err),
+			"ControllerError",
+			fmt.Sprintf("Failed to list StatefulSet pods: %s", err),
 		)
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 	}
 
 	runningPods := 0
 	for _, e := range podList.Items {
-		if e.Status.Phase == "Running" {
+		if resources.PodIsReady(e) {
 			runningPods++
 		}
 	}
@@ -259,7 +261,7 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			databaseNodeSet,
 			corev1.EventTypeNormal,
 			string(DatabaseNodeSetProvisioning),
-			fmt.Sprintf("Waiting for number of running databaseNodeSet pods to match expected: %d != %d", runningPods, databaseNodeSet.Spec.Nodes),
+			fmt.Sprintf("Waiting for number of running nodes to match expected: %d != %d", runningPods, databaseNodeSet.Spec.Nodes),
 		)
 		meta.SetStatusCondition(&databaseNodeSet.Status.Conditions, metav1.Condition{
 			Type:    NodeSetProvisionedCondition,
@@ -275,7 +277,7 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			Type:    NodeSetProvisionedCondition,
 			Status:  metav1.ConditionTrue,
 			Reason:  ReasonCompleted,
-			Message: fmt.Sprintf("Scaled DatabaseNodeSet to %d successfully", databaseNodeSet.Spec.Nodes),
+			Message: fmt.Sprintf("Successfully scaled to desired number of nodes: %d", databaseNodeSet.Spec.Nodes),
 		})
 		return r.updateStatus(ctx, databaseNodeSet, StatusUpdateRequeueDelay)
 	}

internal/controllers/storage/controller_test.go

@@ -4,7 +4,6 @@ import (
 	"context"
 	"fmt"
 	"path/filepath"
-	"strconv"
 	"strings"
 	"testing"
 
@@ -114,8 +113,7 @@ var _ = Describe("Storage controller medium tests", func() {
 		}
 		Expect(foundVolume).To(BeTrue())
 
-		By("Check that label and annotation propagated to pods...", func() {
-			podLabels := storageSS.Spec.Template.Labels
+		By("Check that configuration checksum annotation propagated to pods...", func() {
 			podAnnotations := storageSS.Spec.Template.Annotations
 
 			foundStorage := v1alpha1.Storage{}
@@ -124,12 +122,6 @@ var _ = Describe("Storage controller medium tests", func() {
 				Namespace: testobjects.YdbNamespace,
 			}, &foundStorage)).Should(Succeed())
 
-			foundStorageGenerationLabel := false
-			if podLabels[labels.StorageGeneration] == strconv.FormatInt(foundStorage.ObjectMeta.Generation, 10) {
-				foundStorageGenerationLabel = true
-			}
-			Expect(foundStorageGenerationLabel).To(BeTrue())
-
 			foundConfigurationChecksumAnnotation := false
 			if podAnnotations[annotations.ConfigurationChecksum] == resources.GetConfigurationChecksum(foundStorage.Spec.Configuration) {
 				foundConfigurationChecksumAnnotation = true
@@ -150,5 +142,50 @@ var _ = Describe("Storage controller medium tests", func() {
 			Expect(labelArgKey).Should(BeEquivalentTo(v1alpha1.LabelDeploymentKey))
 			Expect(labelArgValue).Should(BeEquivalentTo(v1alpha1.LabelDeploymentValueKubernetes))
 		})
+
+		By("Check that statefulset podTemplate labels remain immutable...", func() {
+			testLabelKey := "ydb-label"
+			testLabelValue := "test"
+			By("set additional labels to Storage...")
+			Eventually(func() error {
+				foundStorage := v1alpha1.Storage{}
+				Expect(k8sClient.Get(ctx, types.NamespacedName{
+					Name:      storageSample.Name,
+					Namespace: testobjects.YdbNamespace,
+				}, &foundStorage))
+				additionalLabels := resources.CopyDict(foundStorage.Spec.AdditionalLabels)
+				additionalLabels[testLabelKey] = testLabelValue
+				foundStorage.Spec.AdditionalLabels = additionalLabels
+				return k8sClient.Update(ctx, &foundStorage)
+			}, test.Timeout, test.Interval).ShouldNot(HaveOccurred())
+
+			By("check that additional labels was added...")
+			foundStatefulSets := appsv1.StatefulSetList{}
+			Eventually(func() error {
+				err := k8sClient.List(ctx, &foundStatefulSets,
+					client.InNamespace(testobjects.YdbNamespace),
+				)
+				if err != nil {
+					return err
+				}
+				value, exist := foundStatefulSets.Items[0].Labels[testLabelKey]
+				if !exist {
+					return fmt.Errorf("label key `ydb-label` does not exist in StatefulSet. Current labels: %s", foundStatefulSets.Items[0].Labels)
+				}
+				if value != testLabelValue {
+					return fmt.Errorf("label value `ydb-label` in StatefulSet does not equal `test`. Current labels: %s", foundStatefulSets.Items[0].Labels)
+				}
+				return nil
+			}, test.Timeout, test.Interval).ShouldNot(HaveOccurred())
+
+			By("check that StatefulSet selector was not updated...")
+			Expect(*foundStatefulSets.Items[0].Spec.Selector).Should(BeEquivalentTo(
+				metav1.LabelSelector{
+					MatchLabels: map[string]string{
+						labels.StatefulsetComponent: storageSample.Name,
+					},
+				},
+			))
+		})
 	})
 })

internal/controllers/storage/sync.go

@@ -142,42 +142,32 @@ func (r *Reconciler) waitForStatefulSetToScale(
 		r.Recorder.Event(
 			storage,
 			corev1.EventTypeWarning,
-			"ProvisioningFailed",
-			fmt.Sprintf("Failed to get StatefulSets: %s", err),
+			"ControllerError",
+			fmt.Sprintf("Failed to get StatefulSet: %s", err),
 		)
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 	}
 
-	podLabels := labels.Common(storage.Name, make(map[string]string))
-	podLabels.Merge(map[string]string{
-		labels.ComponentKey: labels.StorageComponent,
-	})
-
-	matchingLabels := client.MatchingLabels{}
-	for k, v := range podLabels {
-		matchingLabels[k] = v
-	}
-
 	podList := &corev1.PodList{}
 	opts := []client.ListOption{
 		client.InNamespace(storage.Namespace),
-		matchingLabels,
+		client.MatchingLabels{labels.StatefulsetComponent: storage.Name},
 	}
 
 	err = r.List(ctx, podList, opts...)
 	if err != nil {
 		r.Recorder.Event(
 			storage,
 			corev1.EventTypeWarning,
-			"ProvisioningFailed",
-			fmt.Sprintf("Failed to list cluster pods: %s", err),
+			"ControllerError",
+			fmt.Sprintf("Failed to list StatefulSet pods: %s", err),
 		)
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 	}
 
 	runningPods := 0
 	for _, e := range podList.Items {
-		if e.Status.Phase == "Running" {
+		if resources.PodIsReady(e) {
 			runningPods++
 		}
 	}
@@ -187,7 +177,7 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			storage,
 			corev1.EventTypeNormal,
 			string(StorageProvisioning),
-			fmt.Sprintf("Waiting for number of running storage pods to match expected: %d != %d", runningPods, storage.Spec.Nodes),
+			fmt.Sprintf("Waiting for number of running nodes to match expected: %d != %d", runningPods, storage.Spec.Nodes),
 		)
 		meta.SetStatusCondition(&storage.Status.Conditions, metav1.Condition{
 			Type:    StorageProvisionedCondition,
@@ -203,7 +193,7 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			Type:    StorageProvisionedCondition,
 			Status:  metav1.ConditionTrue,
 			Reason:  ReasonCompleted,
-			Message: "Successfully scaled to desired number of nodes",
+			Message: fmt.Sprintf("Successfully scaled to desired number of nodes: %d", storage.Spec.Nodes),
 		})
 		return r.updateStatus(ctx, storage, StatusUpdateRequeueDelay)
 	}

internal/controllers/storagenodeset/sync.go

@@ -19,6 +19,7 @@ import (
 
 	"github.com/ydb-platform/ydb-kubernetes-operator/api/v1alpha1"
 	. "github.com/ydb-platform/ydb-kubernetes-operator/internal/controllers/constants" //nolint:revive,stylecheck
+	"github.com/ydb-platform/ydb-kubernetes-operator/internal/labels"
 	"github.com/ydb-platform/ydb-kubernetes-operator/internal/resources"
 )
 
@@ -213,49 +214,40 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			r.Recorder.Event(
 				storageNodeSet,
 				corev1.EventTypeWarning,
-				"Syncing",
-				fmt.Sprintf("Failed to found StatefulSet: %s", err),
+				"ProvisioningFailed",
+				fmt.Sprintf("StatefulSet with name %s was not found: %s", storageNodeSet.Name, err),
 			)
 			return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 		}
 		r.Recorder.Event(
 			storageNodeSet,
 			corev1.EventTypeWarning,
-			"Syncing",
-			fmt.Sprintf("Failed to get StatefulSets: %s", err),
+			"ControllerError",
+			fmt.Sprintf("Failed to get StatefulSet: %s", err),
 		)
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 	}
 
-	matchingLabels := client.MatchingLabels{}
-	for k, v := range storageNodeSet.Labels {
-		matchingLabels[k] = v
-	}
-
 	podList := &corev1.PodList{}
 	opts := []client.ListOption{
 		client.InNamespace(storageNodeSet.Namespace),
-		matchingLabels,
+		client.MatchingLabels{labels.StatefulsetComponent: storageNodeSet.Name},
 	}
-	if err = r.List(ctx, podList, opts...); err != nil {
+
+	err = r.List(ctx, podList, opts...)
+	if err != nil {
 		r.Recorder.Event(
 			storageNodeSet,
 			corev1.EventTypeWarning,
-			"Syncing",
-			fmt.Sprintf("Failed to list storageNodeSet pods: %s", err),
+			"ControllerError",
+			fmt.Sprintf("Failed to list StatefulSet pods: %s", err),
 		)
-		meta.SetStatusCondition(&storageNodeSet.Status.Conditions, metav1.Condition{
-			Type:    NodeSetProvisionedCondition,
-			Status:  metav1.ConditionFalse,
-			Reason:  ReasonInProgress,
-			Message: "Failed to check Pods .status.phase",
-		})
 		return Stop, ctrl.Result{RequeueAfter: DefaultRequeueDelay}, err
 	}
 
 	runningPods := 0
 	for _, e := range podList.Items {
-		if e.Status.Phase == "Running" {
+		if resources.PodIsReady(e) {
 			runningPods++
 		}
 	}
@@ -265,7 +257,7 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			storageNodeSet,
 			corev1.EventTypeNormal,
 			string(StorageNodeSetProvisioning),
-			fmt.Sprintf("Waiting for number of running storageNodeSet pods to match expected: %d != %d", runningPods, storageNodeSet.Spec.Nodes),
+			fmt.Sprintf("Waiting for number of running nodes to match expected: %d != %d", runningPods, storageNodeSet.Spec.Nodes),
 		)
 		meta.SetStatusCondition(&storageNodeSet.Status.Conditions, metav1.Condition{
 			Type:    NodeSetProvisionedCondition,
@@ -281,7 +273,7 @@ func (r *Reconciler) waitForStatefulSetToScale(
 			Type:    NodeSetProvisionedCondition,
 			Status:  metav1.ConditionTrue,
 			Reason:  ReasonCompleted,
-			Message: fmt.Sprintf("Scaled StorageNodeSet to %d successfully", storageNodeSet.Spec.Nodes),
+			Message: fmt.Sprintf("Successfully scaled to desired number of nodes: %d", storageNodeSet.Spec.Nodes),
 		})
 		return r.updateStatus(ctx, storageNodeSet, StatusUpdateRequeueDelay)
 	}