Add topology aware deletion priority example

apis/v1alpha1/gameserverset_types.go

@@ -40,6 +40,20 @@ const (
 	InplaceUpdateNotReadyBlocker = "game.kruise.io/inplace-update-not-ready-blocker"
 )
 
+type TopologyDeletionPriorityConfig struct {
+	// +optional
+	// +kubebuilder:default=100
+	BasePriority int `json:"basePriority,omitempty"`
+
+	// +optional
+	// +kubebuilder:default=10
+	PodCountWeight int `json:"podCountWeight,omitempty"`
+
+	// +optional
+	// +kubebuilder:default=5
+	OwnerCountWeight int `json:"ownerCountWeight,omitempty"`
+}
+
 // GameServerSetSpec defines the desired state of GameServerSet
 type GameServerSetSpec struct {
 	// replicas is the desired number of replicas of the given Template.
@@ -54,6 +68,9 @@ type GameServerSetSpec struct {
 	ServiceName          string               `json:"serviceName,omitempty"`
 	ReserveGameServerIds []intstr.IntOrString `json:"reserveGameServerIds,omitempty"`
 	ServiceQualities     []ServiceQuality     `json:"serviceQualities,omitempty"`
+	// +optional
+	TopologyDeletionPriority *TopologyDeletionPriorityConfig `json:"topologyDeletionPriority,omitempty"`
+
 	UpdateStrategy       UpdateStrategy       `json:"updateStrategy,omitempty"`
 	ScaleStrategy        ScaleStrategy        `json:"scaleStrategy,omitempty"`
 	Network              *Network             `json:"network,omitempty"`

config/crd/bases/game.kruise.io_gameserversets.yaml

@@ -21,7 +21,7 @@ spec:
       jsonPath: .spec.replicas
       name: DESIRED
       type: integer
-    - description: The number of currently all GameServers.
+    - description: The total number of current GameServers.
       jsonPath: .status.currentReplicas
       name: CURRENT
       type: integer
@@ -794,7 +794,7 @@ spec:
                                 image:
                                   description: |-
                                     Image indicates the image of the container to update.
-                                    When Image updated, pod.spec.containers[*].image will be updated immediately.
+                                    When Image is updated, pod.spec.containers[*].image will be updated immediately.
                                   type: string
                                 name:
                                   description: Name indicates the name of the container
@@ -803,8 +803,8 @@ spec:
                                 resources:
                                   description: |-
                                     Resources indicates the resources of the container to update.
-                                    When Resources updated, pod.spec.containers[*].Resources will be not updated immediately,
-                                    which will be updated when pod recreate.
+                                    When Resources are updated, pod.spec.containers[*].Resources will not be updated immediately,
+                                    which will be updated when the pod is recreated.
                                   properties:
                                     claims:
                                       description: |-
@@ -882,8 +882,8 @@ spec:
                             type: string
                           result:
                             description: |-
-                              Result indicate the probe message returned by the script.
-                              When Result is defined, it would exec action only when the according Result is actually returns.
+                              Result indicates the probe message returned by the script.
+                              When Result is defined, it would exec action only when the corresponding Result is actually returned.
                             type: string
                           state:
                             type: boolean
@@ -947,6 +947,18 @@ spec:
                   - permanent
                   type: object
                 type: array
+              topologyDeletionPriority:
+                properties:
+                  basePriority:
+                    default: 100
+                    type: integer
+                  ownerCountWeight:
+                    default: 5
+                    type: integer
+                  podCountWeight:
+                    default: 10
+                    type: integer
+                type: object
               updateStrategy:
                 properties:
                   rollingUpdate:
@@ -957,7 +969,7 @@ spec:
                         description: |-
                           UnorderedUpdate contains strategies for non-ordered update.
                           If it is not nil, pods will be updated with non-ordered sequence.
-                          Noted that UnorderedUpdate can only be allowed to work with Parallel podManagementPolicy
+                          Note that UnorderedUpdate can only be allowed to work with Parallel podManagementPolicy
                           UnorderedUpdate *kruiseV1beta1.UnorderedUpdateStrategy `json:"unorderedUpdate,omitempty"`
                           InPlaceUpdateStrategy contains strategies for in-place update.
                         properties:

examples/topology-deletion-priority-automated/gameserverset.yaml

@@ -0,0 +1,21 @@
+apiVersion: game.kruise.io/v1alpha1
+kind: GameServerSet
+metadata:
+  name: gs-topology-automated
+  namespace: default
+spec:
+  replicas: 10
+  topologyDeletionPriority:
+    basePriority: 100
+    podCountWeight: 10
+    ownerCountWeight: 5
+
+  gameServerTemplate:
+    spec:
+      containers:
+      - name: minecraft
+        image: minecraft:latest
+        resources:
+          requests:
+            cpu: 100m
+            memory: 128Mi

pkg/controllers/gameserver/gameserver_manager.go

@@ -73,6 +73,7 @@ type GameServerManager struct {
 	client        client.Client
 	eventRecorder record.EventRecorder
 	logger        logr.Logger
+	topologyRater *TopologyRater
 }
 
 func isNeedToSyncMetadata(gss *gameKruiseV1alpha1.GameServerSet, gs *gameKruiseV1alpha1.GameServer) bool {
@@ -383,13 +384,30 @@ func (manager GameServerManager) SyncPodToGs(ctx context.Context, gss *gameKruis
 		return err
 	}
 
-	// patch gs status
+	if gss.Spec.TopologyDeletionPriority != nil && gs.Spec.DeletionPriority == nil {
+		if manager.topologyRater != nil {
+			topologyPriority, err := manager.topologyRater.CalculateDeletionPriority(ctx, gs, pod, gss.Spec.TopologyDeletionPriority)
+			if err != nil {
+				manager.logger.Error(err, "failed to calculate topology-based deletion priority",
+					telemetryfields.FieldGameServerNamespace, gs.GetNamespace(),
+					telemetryfields.FieldGameServerName, gs.GetName())
+			} else if topologyPriority != nil {
+				podDeletePriority = *topologyPriority
+				manager.logger.V(1).Info("calculated topology-based deletion priority",
+					telemetryfields.FieldGameServerNamespace, gs.GetNamespace(),
+					telemetryfields.FieldGameServerName, gs.GetName(),
+					"priority", topologyPriority.String())
+			}
+		}
+	}
+
 	newStatus := gameKruiseV1alpha1.GameServerStatus{
 		PodStatus:                 pod.Status,
 		CurrentState:              podGsState,
 		DesiredState:              gameKruiseV1alpha1.Ready,
 		UpdatePriority:            &podUpdatePriority,
 		DeletionPriority:          &podDeletePriority,
+
 		ServiceQualitiesCondition: sqConditions,
 		NetworkStatus:             manager.syncNetworkStatus(),
 		LastTransitionTime:        oldGsStatus.LastTransitionTime,
@@ -619,12 +637,13 @@ func (manager GameServerManager) syncPodContainers(gsContainers []gameKruiseV1al
 	return newContainers
 }
 
-func NewGameServerManager(gs *gameKruiseV1alpha1.GameServer, pod *corev1.Pod, c client.Client, recorder record.EventRecorder, logger logr.Logger) Control {
+func NewGameServerManager(gameServer *gameKruiseV1alpha1.GameServer, pod *corev1.Pod, c client.Client, eventRecorder record.EventRecorder, logger logr.Logger) Control {
 	return &GameServerManager{
-		gameServer:    gs,
+		gameServer:    gameServer,
 		pod:           pod,
 		client:        c,
-		eventRecorder: recorder,
+		eventRecorder: eventRecorder,
 		logger:        logger,
+		topologyRater: NewTopologyRater(c),
 	}
 }

pkg/controllers/gameserver/gameserver_manager_test.go

@@ -1208,3 +1208,4 @@ func TestSyncPodToGs(t *testing.T) {
 		}
 	}
 }
+

pkg/controllers/gameserver/topology_rater.go

@@ -0,0 +1,88 @@
+package gameserver
+
+import (
+	"context"
+	"fmt"
+
+	corev1 "k8s.io/api/core/v1"
+	"k8s.io/apimachinery/pkg/fields"
+	"k8s.io/apimachinery/pkg/types"
+	"k8s.io/apimachinery/pkg/util/intstr"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+
+	gameKruiseV1alpha1 "github.com/openkruise/kruise-game/apis/v1alpha1"
+)
+
+const (
+	defaultBasePriority    = 100
+	defaultPodCountWeight  = 10
+	defaultOwnerCountWeight = 5
+)
+
+type TopologyRater struct {
+	client client.Client
+}
+
+func NewTopologyRater(c client.Client) *TopologyRater {
+	return &TopologyRater{client: c}
+}
+
+func (r *TopologyRater) CalculateDeletionPriority(
+	ctx context.Context,
+	gs *gameKruiseV1alpha1.GameServer,
+	pod *corev1.Pod,
+	config *gameKruiseV1alpha1.TopologyDeletionPriorityConfig,
+) (*intstr.IntOrString, error) {
+	if config == nil {
+		return nil, nil
+	}
+
+	nodeName := pod.Spec.NodeName
+	if nodeName == "" {
+		return nil, nil
+	}
+
+	basePriority := defaultBasePriority
+	podCountWeight := defaultPodCountWeight
+	ownerCountWeight := defaultOwnerCountWeight
+
+	if config.BasePriority != 0 {
+		basePriority = config.BasePriority
+	}
+	if config.PodCountWeight != 0 {
+		podCountWeight = config.PodCountWeight
+	}
+	if config.OwnerCountWeight != 0 {
+		ownerCountWeight = config.OwnerCountWeight
+	}
+
+	podList := &corev1.PodList{}
+	listOpts := &client.ListOptions{
+		Namespace:     pod.Namespace,
+		FieldSelector: fields.SelectorFromSet(fields.Set{"spec.nodeName": nodeName}),
+	}
+
+	if err := r.client.List(ctx, podList, listOpts); err != nil {
+		return nil, fmt.Errorf("failed to list pods on node %s: %w", nodeName, err)
+	}
+
+	podCount := len(podList.Items)
+	ownerSet := make(map[types.UID]struct{})
+
+	for _, p := range podList.Items {
+		ownerRefs := p.GetOwnerReferences()
+		for _, owner := range ownerRefs {
+			ownerSet[owner.UID] = struct{}{}
+		}
+	}
+	ownerCount := len(ownerSet)
+
+	priority := basePriority - (podCount * podCountWeight) - (ownerCount * ownerCountWeight)
+
+	if priority < 0 {
+		priority = 0
+	}
+
+	result := intstr.FromInt(priority)
+	return &result, nil
+}

pkg/controllers/gameserver/topology_rater_test.go

@@ -0,0 +1,62 @@
+package gameserver
+
+import (
+	"context"
+	"testing"
+
+	gameKruiseV1alpha1 "github.com/openkruise/kruise-game/apis/v1alpha1"
+	corev1 "k8s.io/api/core/v1"
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/runtime"
+	"k8s.io/apimachinery/pkg/types"
+	"sigs.k8s.io/controller-runtime/pkg/client"
+	"sigs.k8s.io/controller-runtime/pkg/client/fake"
+)
+
+// TestTopologyRater_Basic covers the main happy-path calculation using defaults.
+func TestTopologyRater_Basic(t *testing.T) {
+	ctx := context.TODO()
+
+	scheme := runtime.NewScheme()
+	_ = corev1.AddToScheme(scheme)
+	_ = gameKruiseV1alpha1.AddToScheme(scheme)
+
+	pod := &corev1.Pod{
+		ObjectMeta: metav1.ObjectMeta{
+			Namespace: "default",
+			Name:      "pod-1",
+			OwnerReferences: []metav1.OwnerReference{
+				{UID: types.UID("owner-1")},
+			},
+		},
+		Spec: corev1.PodSpec{
+			NodeName: "node-a",
+		},
+	}
+
+	cl := fake.NewClientBuilder().
+		WithScheme(scheme).
+		WithObjects(pod).
+		WithIndex(&corev1.Pod{}, "spec.nodeName", func(obj client.Object) []string {
+			pod := obj.(*corev1.Pod)
+			return []string{pod.Spec.NodeName}
+		}).
+		Build()
+	rater := NewTopologyRater(cl)
+
+	cfg := &gameKruiseV1alpha1.TopologyDeletionPriorityConfig{}
+
+	got, err := rater.CalculateDeletionPriority(ctx, &gameKruiseV1alpha1.GameServer{}, pod, cfg)
+	if err != nil {
+		t.Fatalf("CalculateDeletionPriority returned error: %v", err)
+	}
+	if got == nil {
+		t.Fatalf("expected non-nil priority")
+	}
+
+	// With one pod and one owner on the node and default weights:
+	// priority = 100 - 1*10 - 1*5 = 85.
+	if got.IntVal != 85 {
+		t.Fatalf("unexpected priority: got=%d want=%d", got.IntVal, 85)
+	}
+}