use simulated render for e2e test

Signed-off-by: Maya Barnea <mayab@il.ibm.com>

Author: mayabar

deploy/environments/dev/base-kind-istio/kustomization.yaml

@@ -16,6 +16,7 @@ kind: Kustomization
 resources:
 - destination-rules.yaml
 - services.yaml
+- vllm-render.yaml
 - ../../../components/istio-control-plane/
 - ../../../components/inference-gateway/
 

deploy/environments/dev/base-kind-istio/vllm-render.yaml

@@ -0,0 +1,69 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: vllm-render
+  labels:
+    app: vllm-render
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: vllm-render
+  template:
+    metadata:
+      labels:
+        app: vllm-render
+    spec:
+      containers:
+      - name: vllm-render
+        image: ${VLLM_RENDER_IMAGE}
+        imagePullPolicy: IfNotPresent
+        command: ["vllm", "launch", "render"]
+        args: ["${MODEL_NAME}", "--port=8082"]
+        ports:
+        - name: render-http
+          containerPort: 8082
+        startupProbe:
+          httpGet:
+            path: /health
+            port: 8082
+          initialDelaySeconds: 2
+          periodSeconds: 1
+          failureThreshold: 600
+          timeoutSeconds: 1
+        livenessProbe:
+          httpGet:
+            path: /health
+            port: 8082
+          periodSeconds: 10
+          failureThreshold: 5
+          timeoutSeconds: 1
+        readinessProbe:
+          httpGet:
+            path: /health
+            port: 8082
+          periodSeconds: 5
+          failureThreshold: 1
+          timeoutSeconds: 1
+        volumeMounts:
+        - name: model-cache
+          mountPath: /root/.cache/huggingface
+      volumes: 
+      - name: model-cache
+        emptyDir: {}
+---
+apiVersion: v1
+kind: Service
+metadata:
+  name: vllm-render
+  labels:
+    app: vllm-render
+spec:
+  selector:
+    app: vllm-render
+  ports:
+  - name: http
+    protocol: TCP
+    port: 8082
+    targetPort: 8082
+  type: ClusterIP

deploy/environments/dev/e2e-infra/vllm-render.yaml

@@ -18,11 +18,11 @@ spec:
       - name: vllm-render
         image: ${VLLM_RENDER_IMAGE}
         imagePullPolicy: IfNotPresent
-        command: ["vllm", "launch", "render"]
-        args: ["${MODEL_NAME}", "--port=8082"]
-        env:
-        - name: HF_TOKEN
-          value: "${HF_TOKEN}"
+        args:
+        - "--model=${MODEL_NAME}"
+        - "--port=8082"
+        - "--mode=echo"
+        - "--force-dummy-tokenizer"
         ports:
         - name: http
           containerPort: 8082
@@ -48,12 +48,7 @@ spec:
           periodSeconds: 5
           failureThreshold: 1
           timeoutSeconds: 1
-        volumeMounts:
-        - name: model-cache
-          mountPath: /root/.cache/huggingface
-      volumes:
-      - name: model-cache
-        emptyDir: {}
+      volumes: []
 ---
 apiVersion: v1
 kind: Service

test/e2e/e2e_suite_test.go

@@ -74,10 +74,10 @@ var (
 
 	containerRuntime = env.GetEnvString("CONTAINER_RUNTIME", "docker", ginkgo.GinkgoLogr)
 	eppImage         = env.GetEnvString("EPP_IMAGE", "ghcr.io/llm-d/llm-d-router-endpoint-picker:dev", ginkgo.GinkgoLogr)
-	vllmSimImage     = env.GetEnvString("VLLM_IMAGE", "ghcr.io/llm-d/llm-d-inference-sim:v0.9.2", ginkgo.GinkgoLogr)
-	sideCarImage     = env.GetEnvString("SIDECAR_IMAGE", "ghcr.io/llm-d/llm-d-router-disagg-sidecar:dev", ginkgo.GinkgoLogr)
-	vllmRenderImage  = env.GetEnvString("VLLM_RENDER_IMAGE", "vllm/vllm-openai-cpu:v0.21.0", ginkgo.GinkgoLogr)
-	loadRenderImage  = env.GetEnvBool("LOAD_VLLM_RENDER_IMAGE", true, ginkgo.GinkgoLogr)
+	vllmSimImage    = env.GetEnvString("VLLM_IMAGE", "ghcr.io/llm-d/llm-d-inference-sim:v0.9.2", ginkgo.GinkgoLogr)
+	sideCarImage    = env.GetEnvString("SIDECAR_IMAGE", "ghcr.io/llm-d/llm-d-router-disagg-sidecar:dev", ginkgo.GinkgoLogr)
+	vllmRenderImage = env.GetEnvString("VLLM_RENDER_IMAGE", "ghcr.io/llm-d/llm-d-inference-sim:v0.9.2", ginkgo.GinkgoLogr)
+	loadRenderImage = env.GetEnvBool("LOAD_VLLM_RENDER_IMAGE", true, ginkgo.GinkgoLogr)
 	// nsName is the namespace in which the K8S objects will be created
 	nsName = env.GetEnvString("NAMESPACE", "default", ginkgo.GinkgoLogr)
 
@@ -212,7 +212,7 @@ func setupK8sCluster() {
 	kindLoadImage(vllmSimImage)
 	kindLoadImage(eppImage)
 	kindLoadImage(sideCarImage)
-	if loadRenderImage {
+	if loadRenderImage && vllmRenderImage != vllmSimImage {
 		kindLoadImage(vllmRenderImage)
 	}
 }

test/e2e/setup_test.go

@@ -28,9 +28,9 @@ func createModelServersFromKustomize(kustomizeDir string, extra map[string]strin
 		"${DECODE_ROLE}":             "",
 		"${EPP_NAME}":                "e2e-epp",
 		"${NAMESPACE}":               nsName,
-		"${VLLM_EXTRA_ARGS_E}":       "",
-		"${VLLM_EXTRA_ARGS_P}":       "",
-		"${VLLM_EXTRA_ARGS_D}":       "",
+		"${VLLM_EXTRA_ARGS_E}":       "--force-dummy-tokenizer",
+		"${VLLM_EXTRA_ARGS_P}":       "--force-dummy-tokenizer",
+		"${VLLM_EXTRA_ARGS_D}":       "--force-dummy-tokenizer",
 	}
 	for k, v := range extra {
 		subs[k] = v
@@ -124,7 +124,6 @@ func createRender() []string {
 		map[string]string{
 			"${MODEL_NAME}":        kvModelName,
 			"${VLLM_RENDER_IMAGE}": vllmRenderImage,
-			"${HF_TOKEN}":          "",
 		})
 	objects := testutils.CreateObjsFromYaml(testConfig, renderYamls)
 	podsInDeploymentsReady(objects)

test/scripts/run_e2e.sh

@@ -21,6 +21,8 @@ echo "Running end to end tests"
 
 DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
 
+export VLLM_RENDER_IMAGE=${VLLM_IMAGE}
+
 if [ -n "${E2E_LABEL_FILTER:-}" ]; then
   echo "Label filter: ${E2E_LABEL_FILTER}"
   go test -v -timeout 45m "${DIR}/../e2e/" -ginkgo.v -ginkgo.fail-fast "-ginkgo.label-filter=${E2E_LABEL_FILTER}"