Addressed first round of comments

Author: ahg-g

pkg/ext-proc/backend/datastore.go

@@ -26,12 +26,12 @@ type Datastore interface {
 
 	// InferenceModel operations
 	ModelSet(infModel *v1alpha1.InferenceModel)
-	ModelGet(modelName string) (returnModel *v1alpha1.InferenceModel)
+	ModelGet(modelName string) (*v1alpha1.InferenceModel, bool)
 	ModelDelete(modelName string)
 
 	// PodMetrics operations
 	PodUpdateOrAddIfNotExist(pod *corev1.Pod) bool
-	PodUpdateMetricsIfExist(pm *PodMetrics)
+	PodUpdateMetricsIfExist(pm *PodMetrics) bool
 	PodGet(namespacedName types.NamespacedName) (*PodMetrics, bool)
 	PodDelete(namespacedName types.NamespacedName)
 	PodResyncAll(ctx context.Context, ctrlClient client.Client)
@@ -102,24 +102,26 @@ func (ds *datastore) ModelSet(infModel *v1alpha1.InferenceModel) {
 	ds.models.Store(infModel.Spec.ModelName, infModel)
 }
 
-func (ds *datastore) ModelGet(modelName string) (returnModel *v1alpha1.InferenceModel) {
+func (ds *datastore) ModelGet(modelName string) (*v1alpha1.InferenceModel, bool) {
 	infModel, ok := ds.models.Load(modelName)
 	if ok {
-		returnModel = infModel.(*v1alpha1.InferenceModel)
+		return infModel.(*v1alpha1.InferenceModel), true
 	}
-	return
+	return nil, false
 }
 
 func (ds *datastore) ModelDelete(modelName string) {
 	ds.models.Delete(modelName)
 }
 
 // /// Pods/endpoints APIs ///
-func (ds *datastore) PodUpdateMetricsIfExist(pm *PodMetrics) {
+func (ds *datastore) PodUpdateMetricsIfExist(pm *PodMetrics) bool {
 	if val, ok := ds.pods.Load(pm.NamespacedName); ok {
 		existing := val.(*PodMetrics)
 		existing.Metrics = pm.Metrics
+		return true
 	}
+	return false
 }
 
 func (ds *datastore) PodGet(namespacedName types.NamespacedName) (*PodMetrics, bool) {

pkg/ext-proc/backend/inferencemodel_reconciler_test.go

@@ -242,7 +242,7 @@ func TestReconcile_ModelMarkedForDeletion(t *testing.T) {
 	}
 
 	// Verify that the datastore was not updated.
-	if infModel := datastore.ModelGet(existingModel.Spec.ModelName); infModel != nil {
+	if _, exist := datastore.ModelGet(existingModel.Spec.ModelName); exist {
 		t.Errorf("expected datastore to not contain model %q", existingModel.Spec.ModelName)
 	}
 }
@@ -299,7 +299,7 @@ func TestReconcile_ResourceExists(t *testing.T) {
 	}
 
 	// Verify that the datastore was updated.
-	if infModel := datastore.ModelGet(existingModel.Spec.ModelName); infModel == nil {
+	if _, exist := datastore.ModelGet(existingModel.Spec.ModelName); !exist {
 		t.Errorf("expected datastore to contain model %q", existingModel.Spec.ModelName)
 	}
 }

pkg/ext-proc/handlers/request.go

@@ -48,8 +48,8 @@ func (s *Server) HandleRequestBody(
 	// NOTE: The nil checking for the modelObject means that we DO allow passthrough currently.
 	// This might be a security risk in the future where adapters not registered in the InferenceModel
 	// are able to be requested by using their distinct name.
-	modelObj := s.datastore.ModelGet(model)
-	if modelObj == nil {
+	modelObj, exist := s.datastore.ModelGet(model)
+	if !exist {
 		return nil, fmt.Errorf("error finding a model object in InferenceModel for input %v", model)
 	}
 	if len(modelObj.Spec.TargetModels) > 0 {

pkg/manifests/ext_proc.yaml

@@ -71,7 +71,7 @@ spec:
     spec:
       containers:
       - name: inference-gateway-ext-proc
-        image: us-central1-docker.pkg.dev/ahg-gke-dev/jobset2/epp:dfee85a
+        image: us-central1-docker.pkg.dev/k8s-staging-images/gateway-api-inference-extension/epp:main
         imagePullPolicy: Always
         args:
         - -poolName

pkg/manifests/vllm/deployment.yaml

@@ -39,7 +39,7 @@ spec:
           - "8000"
           - "--enable-lora"
           - "--max-loras"
-          - "2"
+          - "4"
           - "--max-cpu-loras"
           - "12"
           - "--lora-modules"

test/integration/hermetic_test.go

@@ -490,7 +490,8 @@ func BeforeSuit(t *testing.T) func() {
 	}
 
 	assert.EventuallyWithT(t, func(t *assert.CollectT) {
-		synced := serverRunner.Datastore.PoolHasSynced() && serverRunner.Datastore.ModelGet("my-model") != nil
+		_, modelExist := serverRunner.Datastore.ModelGet("my-model")
+		synced := serverRunner.Datastore.PoolHasSynced() && modelExist
 		assert.True(t, synced, "Timeout waiting for the pool and models to sync")
 	}, 10*time.Second, 10*time.Millisecond)