add verifyenv

Signed-off-by: GitHub <[email protected]>

Author: samzong

e2e/profiles/llm-d/profile.go

@@ -244,6 +244,55 @@ func (p *Profile) verifyEnvironment(ctx context.Context, opts *framework.SetupOp
 		return err
 	}
 
+	// Verify required CRDs/APIs from Gateway API and Inference Extension are registered.
+	type apiCheck struct {
+		groupVersion      string
+		expectedResources []string
+		optional          bool
+	}
+	checkAPIGroup := func(c apiCheck) error {
+		resources, err := client.Discovery().ServerResourcesForGroupVersion(c.groupVersion)
+		if err != nil {
+			if c.optional {
+				if p.verbose {
+					fmt.Printf("[Verify] API group %s not found (optional): %v\n", c.groupVersion, err)
+				}
+				return nil
+			}
+			return fmt.Errorf("discover %s: %w", c.groupVersion, err)
+		}
+		found := make(map[string]bool, len(resources.APIResources))
+		for _, r := range resources.APIResources {
+			found[r.Name] = true
+		}
+		for _, r := range c.expectedResources {
+			if !found[r] {
+				if c.optional {
+					if p.verbose {
+						fmt.Printf("[Verify] Missing optional resource %s in %s\n", r, c.groupVersion)
+					}
+					return nil
+				}
+				return fmt.Errorf("missing %s in %s", r, c.groupVersion)
+			}
+		}
+		if p.verbose {
+			fmt.Printf("[Verify] API group %s present with %v\n", c.groupVersion, c.expectedResources)
+		}
+		return nil
+	}
+
+	for _, c := range []apiCheck{
+		{groupVersion: "gateway.networking.k8s.io/v1", expectedResources: []string{"gateways", "httproutes"}},
+		{groupVersion: "inference.networking.k8s.io/v1", expectedResources: []string{"inferencepools"}},
+		// EndpointPickerConfig CRD is optional in some environments; treat as best-effort.
+		{groupVersion: "inference.networking.x-k8s.io/v1alpha1", expectedResources: []string{"endpointpickerconfigs"}, optional: true},
+	} {
+		if err := checkAPIGroup(c); err != nil {
+			return err
+		}
+	}
+
 	// Actively wait for critical deployments to become Available before checking readiness counts.
 	// This avoids flakiness when resources are still pulling images just after creation.
 	deployer := helm.NewDeployer(opts.KubeConfig, opts.Verbose)
@@ -252,6 +301,8 @@ func (p *Profile) verifyEnvironment(ctx context.Context, opts *framework.SetupOp
 	}{
 		{semanticNamespace, "semantic-router"},
 		{gatewayNamespace, "istiod"},
+		{"default", "vllm-llama3-8b-instruct"},
+		{"default", "phi4-mini"},
 		{"default", "llm-d-inference-scheduler-llama3-8b"},
 		{"default", "llm-d-inference-scheduler-phi4-mini"},
 		{"default", "inference-gateway-istio"},
@@ -268,6 +319,12 @@ func (p *Profile) verifyEnvironment(ctx context.Context, opts *framework.SetupOp
 	if err := helpers.CheckDeployment(ctx, client, gatewayNamespace, "istiod", p.verbose); err != nil {
 		return err
 	}
+	if err := helpers.CheckDeployment(ctx, client, "default", "vllm-llama3-8b-instruct", p.verbose); err != nil {
+		return err
+	}
+	if err := helpers.CheckDeployment(ctx, client, "default", "phi4-mini", p.verbose); err != nil {
+		return err
+	}
 	if err := helpers.CheckDeployment(ctx, client, "default", "llm-d-inference-scheduler-llama3-8b", p.verbose); err != nil {
 		return err
 	}