feat: Only monitor NATS metrics if using NATS request plane (#4442)

Signed-off-by: Graham King <grahamk@nvidia.com>

lib/runtime/src/runtime.rs

@@ -324,9 +324,9 @@ impl Runtime {
                 tracker.wait_for_completion().await;
             }
 
-            // Phase 3: Now connections will be disconnected to NATS/ETCD by cancelling the main token
+            // Phase 3: Now connections will be disconnected to backend services (e.g. NATS/ETCD) by cancelling the main token
             tracing::info!(
-                "Phase 3: All endpoints ended gracefully. Connections to NATS/ETCD will now be disconnected"
+                "Phase 3: All endpoints ended gracefully. Connections to backend services will now be disconnected"
             );
             main_token.cancel();
         });

lib/runtime/tests/namespace_etcd_path.rs

@@ -43,7 +43,7 @@ async fn test_recursive_namespace_implementation() {
     let runtime = Runtime::from_current().unwrap();
     let config = DistributedConfig {
         store_backend: KeyValueStoreSelect::Memory,
-        nats_config: nats::ClientOptions::default(),
+        nats_config: Some(nats::ClientOptions::default()),
         request_plane: dynamo_runtime::distributed::RequestPlaneMode::default(),
     };
     let distributed_runtime = DistributedRuntime::new(runtime, config).await.unwrap();
@@ -90,7 +90,7 @@ async fn test_multiple_branches_recursive_namespaces() {
     let runtime = Runtime::from_current().unwrap();
     let config = DistributedConfig {
         store_backend: KeyValueStoreSelect::Memory,
-        nats_config: nats::ClientOptions::default(),
+        nats_config: Some(nats::ClientOptions::default()),
         request_plane: dynamo_runtime::distributed::RequestPlaneMode::default(),
     };
     let distributed_runtime = DistributedRuntime::new(runtime, config).await.unwrap();

tests/frontend/grpc/echo_tensor_worker.py

@@ -16,7 +16,6 @@ from dynamo.runtime import DistributedRuntime, dynamo_worker
 @dynamo_worker()
 async def echo_tensor_worker(runtime: DistributedRuntime):
     component = runtime.namespace("tensor").component("echo")
-    await component.create_service()
 
     endpoint = component.endpoint("generate")
 

tests/serve/launch/template_verifier.py

@@ -44,7 +44,6 @@ async def main(runtime: DistributedRuntime):
 
     # Create service
     component = runtime.namespace("test").component("backend")
-    await component.create_service()
     endpoint = component.endpoint("generate")
 
     # Use the existing custom template from fixtures