[WIP] E2E Test Resource Setup for Multiport Pod

Author: RyanRosario

Makefile

@@ -36,6 +36,7 @@ IMAGE_NAME := epp
 IMAGE_REPO ?= $(IMAGE_REGISTRY)/$(IMAGE_NAME)
 IMAGE_TAG ?= $(IMAGE_REPO):$(GIT_TAG)
 PROJECT_DIR := $(shell dirname $(abspath $(lastword $(MAKEFILE_LIST))))
+# PROJECT_DIR := $(shell pwd)
 # The path to the E2E manifest file. It can be overridden by setting the
 # E2E_MANIFEST_PATH environment variable. Note that HF_TOKEN must be set when using the GPU-based manifest.
 E2E_MANIFEST_PATH ?= config/manifests/vllm/sim-deployment.yaml
@@ -153,8 +154,8 @@ test-integration: envtest ## Run integration tests.
 	CGO_ENABLED=1 KUBEBUILDER_ASSETS="$(shell $(ENVTEST) use $(ENVTEST_K8S_VERSION) --bin-dir $(LOCALBIN) -p path)" go test ./test/integration/epp/... -race -coverprofile cover.out
 
 .PHONY: test-e2e
-test-e2e: ## Run end-to-end tests against an existing Kubernetes cluster.
-	MANIFEST_PATH=$(PROJECT_DIR)/$(E2E_MANIFEST_PATH) E2E_IMAGE=$(E2E_IMAGE) USE_KIND=$(E2E_USE_KIND) ./hack/test-e2e.sh
+test-e2e: generate install ## Run end-to-end tests against an existing Kubernetes cluster.
+    MANIFEST_PATH=$(PROJECT_DIR)/$(E2E_MANIFEST_PATH) E2E_IMAGE=$(E2E_IMAGE) USE_KIND=$(E2E_USE_KIND) ./hack/test-e2e.sh
 
 .PHONY: lint
 lint: golangci-lint ## Run golangci-lint linter
@@ -245,16 +246,13 @@ syncer-image-local-push: syncer-image-local-build
 
 .PHONY: syncer-image-build
 syncer-image-build:
-	$ cd $(CURDIR)/tools/dynamic-lora-sidecar && $(IMAGE_BUILD_CMD) -t $(SYNCER_IMAGE_TAG) \
-		--platform=$(PLATFORMS) \
-		--build-arg BASE_IMAGE=$(BASE_IMAGE) \
-		--build-arg BUILDER_IMAGE=$(BUILDER_IMAGE) \
-		$(PUSH) \
-		$(SYNCER_IMAGE_BUILD_EXTRA_OPTS) ./
-
-.PHONY: syncer-image-push
-syncer-image-push: PUSH=--push
-syncer-image-push: syncer-image-build
+	cd $(PROJECT_DIR)/tools/dynamic-lora-sidecar && \
+	$(IMAGE_BUILD_CMD) -t $(SYNCER_IMAGE_TAG) \
+        --platform=$(PLATFORMS) \
+        --build-arg BASE_IMAGE=$(BASE_IMAGE) \
+        --build-arg BUILDER_IMAGE=$(BUILDER_IMAGE) \
+        $(PUSH) \
+        $(SYNCER_IMAGE_BUILD_EXTRA_OPTS) .
 
 ##@ Body-based Routing extension
 

config/manifests/inferenceobjective.yaml

@@ -4,6 +4,8 @@ metadata:
   name: food-review
 spec:
   priority: 1
+  subresources:
+    status: {}
   poolRef:
     group: inference.networking.k8s.io
     name: vllm-llama3-8b-instruct
@@ -14,6 +16,8 @@ metadata:
   name: base-model
 spec:
   priority: 2
+  subresources:
+    status: {}
   poolRef:
     group: inference.networking.k8s.io
     name: vllm-llama3-8b-instruct
@@ -24,6 +28,8 @@ metadata:
   name: base-model-cpu
 spec:
   priority: 2
+  subresources:
+    status: {}
   poolRef:
     group: inference.networking.k8s.io
     name: vllm-llama3-8b-instruct

hack/test-e2e.sh

@@ -34,12 +34,12 @@ if [ "$USE_KIND" = "true" ]; then
   if ! kubectl config current-context >/dev/null 2>&1; then # if no active kind cluster found
     echo "No active kubecontext found. creating a kind cluster for running the tests..."
     kind create cluster --name inference-e2e
-    KIND_CLUSTER=inference-e2e IMAGE_TAG=${E2E_IMAGE} make image-kind
+    KIND_CLUSTER=inference-e2e IMAGE_TAG=${E2E_IMAGE} make s=""  image-kind
   else 
     current_context=$(kubectl config current-context)
     current_kind_cluster="${current_context#kind-}"
     echo "Found an active kind cluster ${current_kind_cluster} for running the tests..."
-    KIND_CLUSTER=${current_kind_cluster} IMAGE_TAG=${E2E_IMAGE} make image-kind
+    KIND_CLUSTER=${current_kind_cluster} IMAGE_TAG=${E2E_IMAGE} make s="" image-kind
   fi 
 else 
   # don't use kind. it's the caller responsibility to load the image into the cluster, we just run the tests.
@@ -51,4 +51,19 @@ else
 fi
 
 echo "Found an active cluster. Running Go e2e tests in ./epp..."
+
+# CRD and Namespace deletion BEFORE THE TEST RUN. 
+# This ensures a clean state for the subsequent 'make install' dependency 
+# (which runs BEFORE this script starts).
+
+kubectl delete namespace inf-ext-e2e --ignore-not-found=true --wait=true
+# Removing the CRD deletion here is safe, as the Makefile now runs 'install' 
+# which depends on 'generate' and 'kustomize build' then 'kubectl apply'.
+
+kubectl delete namespace inf-ext-e2e --ignore-not-found=true --wait=true
+kubectl delete crd inferencepools.inference.networking.k8s.io --ignore-not-found=true --wait=true
+# kubectl delete namespace inf-ext-e2e
+# kubectl delete crd inferencepools.inference.networking.k8s.io
+# kubectl delete inferencepool vllm-llama3-8b-instruct -n inf-ext-e2e
+# MANIFEST_PATH=${E2E_MANIFEST_PATH}
 go test ./test/e2e/epp/ -v -ginkgo.v

image/Dockerfile

@@ -0,0 +1,6 @@
+FROM python:3.12-alpine
+WORKDIR /app
+COPY server.py /app/server.py
+
+# ENTRYPOINT ensures that arguments passed at runtime are appended to this command.
+ENTRYPOINT ["python", "-u", "server.py"]

image/server.py

@@ -0,0 +1,47 @@
+import threading
+import sys
+from http.server import HTTPServer, BaseHTTPRequestHandler
+
+class MultiPortHandler(BaseHTTPRequestHandler):
+    def do_GET(self):
+        self._handle_request()
+
+    # FIX: Add this method to handle the test's POST requests
+    def do_POST(self):
+        # We need to read the body to keep the socket clean, even if we ignore it
+        content_length = int(self.headers.get('Content-Length', 0))
+        self.rfile.read(content_length)
+        self._handle_request()
+
+    def _handle_request(self):
+        self.send_response(200)
+        self.send_header('Content-type', 'text/plain')
+        self.end_headers()
+        response = f"Handled by port: {self.server.server_port}\n"
+        self.wfile.write(response.encode('utf-8'))
+    
+    def log_message(self, format, *args):
+        pass
+
+def start_server(port):
+    print(f"Starting server on port {port}...")
+    server = HTTPServer(('0.0.0.0', port), MultiPortHandler)
+    server.serve_forever()
+
+if __name__ == "__main__":
+    start_port = 8000
+    port_count = 1
+    if len(sys.argv) >= 3:
+        start_port = int(sys.argv[1])
+        port_count = int(sys.argv[2])
+
+    threads = []
+    for i in range(port_count):
+        port = start_port + i
+        thread = threading.Thread(target=start_server, args=(port,))
+        thread.daemon = True
+        thread.start()
+        threads.append(thread)
+
+    for thread in threads:
+        thread.join()

pkg/epp/controller/inferenceobjective_reconciler.go

@@ -27,14 +27,16 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/log"
 	"sigs.k8s.io/controller-runtime/pkg/predicate"
 
+	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
 	"sigs.k8s.io/gateway-api-inference-extension/apix/v1alpha2"
 	"sigs.k8s.io/gateway-api-inference-extension/pkg/common"
 	"sigs.k8s.io/gateway-api-inference-extension/pkg/epp/datastore"
 	logutil "sigs.k8s.io/gateway-api-inference-extension/pkg/epp/util/logging"
 )
 
 type InferenceObjectiveReconciler struct {
-	client.Reader
+	client.Client
 	Datastore datastore.Datastore
 	PoolGKNN  common.GKNN
 }
@@ -65,6 +67,33 @@ func (c *InferenceObjectiveReconciler) Reconcile(ctx context.Context, req ctrl.R
 	c.Datastore.ObjectiveSet(infObjective)
 	logger.Info("Added/Updated InferenceObjective")
 
+	newCondition := metav1.Condition{
+		Type:               "Accepted",
+		Status:             metav1.ConditionTrue, // Use metav1.ConditionTrue in real code
+		Reason:             "PoolReferenceValid",
+		Message:            "InferenceObjective successfully configured and pool reference is valid.",
+		LastTransitionTime: metav1.Now(),
+	}
+
+	// Find and update the status condition list
+	found := false
+	for i, cond := range infObjective.Status.Conditions {
+		if cond.Type == "Accepted" {
+			infObjective.Status.Conditions[i] = newCondition
+			found = true
+			break
+		}
+	}
+	if !found {
+		infObjective.Status.Conditions = append(infObjective.Status.Conditions, newCondition)
+	}
+
+	// Write the status update back to the API server
+	if err := c.Status().Update(ctx, infObjective); err != nil {
+		return ctrl.Result{}, fmt.Errorf("failed to update InferenceObjective status: %w", err)
+	}
+	logger.Info("InferenceObjective Status updated successfully to Accepted: True")
+
 	return ctrl.Result{}, nil
 }
 

pkg/epp/controller/inferenceobjective_reconciler_test.go

@@ -166,7 +166,7 @@ func TestInferenceObjectiveReconciler(t *testing.T) {
 			}
 			_ = ds.PoolSet(context.Background(), fakeClient, pool)
 			reconciler := &InferenceObjectiveReconciler{
-				Reader:    fakeClient,
+				Client:    fakeClient,
 				Datastore: ds,
 				PoolGKNN: common.GKNN{
 					NamespacedName: types.NamespacedName{Name: pool.Name, Namespace: pool.Namespace},

pkg/epp/server/runserver.go

@@ -123,7 +123,7 @@ func (r *ExtProcServerRunner) SetupWithManager(ctx context.Context, mgr ctrl.Man
 
 	if err := (&controller.InferenceObjectiveReconciler{
 		Datastore: r.Datastore,
-		Reader:    mgr.GetClient(),
+		Client:    mgr.GetClient(),
 		PoolGKNN:  r.PoolGKNN,
 	}).SetupWithManager(ctx, mgr); err != nil {
 		return fmt.Errorf("failed setting up InferenceObjectiveReconciler: %w", err)

pkg/epp/server/server_test.go

@@ -34,7 +34,7 @@ import (
 )
 
 const (
-	bufSize    = 1024 * 1024
+	// bufSize    = 1024 * 1024
 	podName    = "pod1"
 	podAddress = "1.2.3.4"
 	poolPort   = int32(5678)

test/e2e/epp/e2e_suite_test.go

@@ -114,10 +114,11 @@ var _ = ginkgo.BeforeSuite(func() {
 	}
 
 	ginkgo.By("Setting up the test suite")
-	setupSuite()
+	setupSuite() // <--- Creates the Client (Initial)
 
 	ginkgo.By("Creating test infrastructure")
-	setupInfra()
+	setupInfra() // <--- Installs the CRDs
+
 })
 
 func setupInfra() {
@@ -127,6 +128,11 @@ func setupInfra() {
 
 	createNamespace(testConfig)
 
+	// Add this section to apply the RBAC AFTER the namespace is created
+	ginkgo.By("Applying InferenceObjective Status RBAC")
+	testutils.ApplyYAMLFile(testConfig, "../../testdata/inferenceobjective-status-rbac.yaml")
+	// gomega.Expect(err).To(gomega.Succeed(), "Failed to apply inferenceobjective-status-rbac.yaml")
+
 	modelServerManifestArray := getYamlsFromModelServerManifest(modelServerManifestPath)
 	if strings.Contains(modelServerManifestArray[0], "hf-token") {
 		createHfSecret(testConfig, modelServerSecretManifest)