diff --git a/vllm/engine/arg_utils.py b/vllm/engine/arg_utils.py
index 119505f08571d..9a02bf83848cb 100644
--- a/vllm/engine/arg_utils.py
+++ b/vllm/engine/arg_utils.py
@@ -1102,6 +1102,7 @@ class EngineArgs:
             placement_group = ray.util.get_current_placement_group()
 
         data_parallel_external_lb = self.data_parallel_rank is not None
+        # Local DP rank = 1, use pure-external LB.
         if data_parallel_external_lb:
             assert self.data_parallel_size_local in (1, None), (
                 "data_parallel_size_local must be 1 when data_parallel_rank "
@@ -1109,6 +1110,7 @@ class EngineArgs:
             data_parallel_size_local = 1
             # Use full external lb if we have local_size of 1.
             self.data_parallel_hybrid_lb = False
+        # Local DP rank > 1, use hybrid LB.
         elif self.data_parallel_hybrid_lb:
             assert self.data_parallel_start_rank is not None, (
                 "data_parallel_start_rank must be set to use "
diff --git a/vllm/entrypoints/cli/serve.py b/vllm/entrypoints/cli/serve.py
index e26642b2f65e4..01f1c1e37155b 100644
--- a/vllm/entrypoints/cli/serve.py
+++ b/vllm/entrypoints/cli/serve.py
@@ -86,8 +86,8 @@ def run_headless(args: argparse.Namespace):
     if not envs.VLLM_USE_V1:
         raise ValueError("Headless mode is only supported for V1")
 
-    if engine_args.data_parallel_rank is not None:
-        raise ValueError("data_parallel_rank is not applicable in "
+    if engine_args.data_parallel_hybrid_lb:
+        raise ValueError("data_parallel_hybrid_lb is not applicable in "
                          "headless mode")
 
     parallel_config = vllm_config.parallel_config