Fix test and fmt

Author: liu-cong

Diff for: pkg/epp/controller/inferencemodel_reconciler_test.go

@@ -192,11 +192,15 @@ func TestInferenceModelReconciler(t *testing.T) {
 				WithIndex(&v1alpha2.InferenceModel{}, datastore.ModelNameIndexKey, indexInferenceModelsByModelName).
 				Build()
 			pmf := backendmetrics.NewPodMetricsFactory(&backendmetrics.FakePodMetricsClient{}, time.Second)
-			datastore := datastore.NewFakeDatastore(t.Context(), pmf, nil, test.modelsInStore, pool)
+			ds := datastore.NewDatastore(t.Context(), pmf)
+			for _, m := range test.modelsInStore {
+				ds.ModelSetIfOlder(m)
+			}
+			ds.PoolSet(pool)
 			reconciler := &InferenceModelReconciler{
 				Client:             fakeClient,
 				Record:             record.NewFakeRecorder(10),
-				Datastore:          datastore,
+				Datastore:          ds,
 				PoolNamespacedName: types.NamespacedName{Name: pool.Name, Namespace: pool.Namespace},
 			}
 			if test.incomingReq == nil {
@@ -213,11 +217,11 @@ func TestInferenceModelReconciler(t *testing.T) {
 				t.Errorf("Unexpected result diff (+got/-want): %s", diff)
 			}
 
-			if len(test.wantModels) != len(datastore.ModelGetAll()) {
-				t.Errorf("Unexpected; want: %d, got:%d", len(test.wantModels), len(datastore.ModelGetAll()))
+			if len(test.wantModels) != len(ds.ModelGetAll()) {
+				t.Errorf("Unexpected; want: %d, got:%d", len(test.wantModels), len(ds.ModelGetAll()))
 			}
 
-			if diff := diffStore(datastore, diffStoreParams{wantPool: pool, wantModels: test.wantModels}); diff != "" {
+			if diff := diffStore(ds, diffStoreParams{wantPool: pool, wantModels: test.wantModels}); diff != "" {
 				t.Errorf("Unexpected diff (+got/-want): %s", diff)
 			}
 

Diff for: pkg/epp/controller/pod_reconciler.go

@@ -59,7 +59,7 @@ func (c *PodReconciler) Reconcile(ctx context.Context, req ctrl.Request) (ctrl.R
 		return ctrl.Result{}, err
 	}
 
-	c.updateDatastore(ctx, logger, pod, pool)
+	c.updateDatastore(logger, pod, pool)
 	return ctrl.Result{}, nil
 }
 
@@ -69,7 +69,7 @@ func (c *PodReconciler) SetupWithManager(mgr ctrl.Manager) error {
 		Complete(c)
 }
 
-func (c *PodReconciler) updateDatastore(ctx context.Context, logger logr.Logger, pod *corev1.Pod, pool *v1alpha2.InferencePool) {
+func (c *PodReconciler) updateDatastore(logger logr.Logger, pod *corev1.Pod, pool *v1alpha2.InferencePool) {
 	namespacedName := types.NamespacedName{Name: pod.Name, Namespace: pod.Namespace}
 	if !pod.DeletionTimestamp.IsZero() || !c.Datastore.PoolLabelsMatch(pod.Labels) || !podIsReady(pod) {
 		logger.V(logutil.DEBUG).Info("Pod removed or not added", "name", namespacedName)

Diff for: test/integration/epp/hermetic_test.go

@@ -95,15 +95,15 @@ func TestKubeInferenceModelRequest(t *testing.T) {
 			req:  utiltesting.GenerateRequest(logger, "test1", "my-model"),
 			// pod-1 will be picked because it has relatively low queue size and low KV cache.
 			pods: map[backendmetrics.Pod]*backendmetrics.Metrics{
-				fakePod(0): &backendmetrics.Metrics{
+				fakePod(0): {
 					WaitingQueueSize:    3,
 					KVCacheUsagePercent: 0.2,
 				},
-				fakePod(1): &backendmetrics.Metrics{
+				fakePod(1): {
 					WaitingQueueSize:    0,
 					KVCacheUsagePercent: 0.1,
 				},
-				fakePod(2): &backendmetrics.Metrics{
+				fakePod(2): {
 					WaitingQueueSize:    10,
 					KVCacheUsagePercent: 0.2,
 				},
@@ -137,23 +137,23 @@ func TestKubeInferenceModelRequest(t *testing.T) {
 			// pod-1 will be picked because it has relatively low queue size, with the requested
 			// model being active, and has low KV cache.
 			pods: map[backendmetrics.Pod]*backendmetrics.Metrics{
-				fakePod(0): &backendmetrics.Metrics{
+				fakePod(0): {
 					WaitingQueueSize:    0,
 					KVCacheUsagePercent: 0.2,
 					ActiveModels: map[string]int{
 						"foo": 1,
 						"bar": 1,
 					},
 				},
-				fakePod(1): &backendmetrics.Metrics{
+				fakePod(1): {
 					WaitingQueueSize:    0,
 					KVCacheUsagePercent: 0.1,
 					ActiveModels: map[string]int{
 						"foo":            1,
 						"sql-lora-1fdg2": 1,
 					},
 				},
-				fakePod(2): &backendmetrics.Metrics{
+				fakePod(2): {
 					WaitingQueueSize:    10,
 					KVCacheUsagePercent: 0.2,
 					ActiveModels: map[string]int{
@@ -192,23 +192,23 @@ func TestKubeInferenceModelRequest(t *testing.T) {
 			// as it's above the affinity for queue size. Also is critical, so we should
 			// still honor request despite all queues > 5
 			pods: map[backendmetrics.Pod]*backendmetrics.Metrics{
-				fakePod(0): &backendmetrics.Metrics{
+				fakePod(0): {
 					WaitingQueueSize:    10,
 					KVCacheUsagePercent: 0.2,
 					ActiveModels: map[string]int{
 						"foo": 1,
 						"bar": 1,
 					},
 				},
-				fakePod(1): &backendmetrics.Metrics{
+				fakePod(1): {
 					WaitingQueueSize:    200,
 					KVCacheUsagePercent: 0.1,
 					ActiveModels: map[string]int{
 						"foo":            1,
 						"sql-lora-1fdg2": 1,
 					},
 				},
-				fakePod(2): &backendmetrics.Metrics{
+				fakePod(2): {
 					WaitingQueueSize:    6,
 					KVCacheUsagePercent: 0.2,
 					ActiveModels: map[string]int{
@@ -245,7 +245,7 @@ func TestKubeInferenceModelRequest(t *testing.T) {
 			// no pods will be picked as all models are either above kv threshold,
 			// queue threshold, or both.
 			pods: map[backendmetrics.Pod]*backendmetrics.Metrics{
-				fakePod(0): &backendmetrics.Metrics{
+				fakePod(0): {
 					WaitingQueueSize:    6,
 					KVCacheUsagePercent: 0.2,
 					ActiveModels: map[string]int{
@@ -254,15 +254,15 @@ func TestKubeInferenceModelRequest(t *testing.T) {
 						"sql-lora-1fdg3": 1,
 					},
 				},
-				fakePod(1): &backendmetrics.Metrics{
+				fakePod(1): {
 					WaitingQueueSize:    0,
 					KVCacheUsagePercent: 0.85,
 					ActiveModels: map[string]int{
 						"foo":            1,
 						"sql-lora-1fdg3": 1,
 					},
 				},
-				fakePod(2): &backendmetrics.Metrics{
+				fakePod(2): {
 					WaitingQueueSize:    10,
 					KVCacheUsagePercent: 0.9,
 					ActiveModels: map[string]int{
@@ -287,7 +287,7 @@ func TestKubeInferenceModelRequest(t *testing.T) {
 			req:  utiltesting.GenerateRequest(logger, "test5", "sql-lora-sheddable"),
 			// pod 0 will be picked as all other models are above threshold
 			pods: map[backendmetrics.Pod]*backendmetrics.Metrics{
-				fakePod(0): &backendmetrics.Metrics{
+				fakePod(0): {
 					WaitingQueueSize:    4,
 					KVCacheUsagePercent: 0.2,
 					ActiveModels: map[string]int{
@@ -296,15 +296,15 @@ func TestKubeInferenceModelRequest(t *testing.T) {
 						"sql-lora-1fdg3": 1,
 					},
 				},
-				fakePod(1): &backendmetrics.Metrics{
+				fakePod(1): {
 					WaitingQueueSize:    0,
 					KVCacheUsagePercent: 0.85,
 					ActiveModels: map[string]int{
 						"foo":            1,
 						"sql-lora-1fdg3": 1,
 					},
 				},
-				fakePod(2): &backendmetrics.Metrics{
+				fakePod(2): {
 					WaitingQueueSize:    10,
 					KVCacheUsagePercent: 0.9,
 					ActiveModels: map[string]int{