[Perf] API-server scaleout with many-to-many server-engine comms (#17546)

tests/utils.py

@@ -28,7 +28,7 @@ from tests.models.utils import TextTextLogprobs
 from vllm.distributed import (ensure_model_parallel_initialized,
                               init_distributed_environment)
 from vllm.engine.arg_utils import AsyncEngineArgs
-from vllm.entrypoints.openai.cli_args import make_arg_parser
+from vllm.entrypoints.cli.serve import ServeSubcommand
 from vllm.model_executor.model_loader import get_model_loader
 from vllm.platforms import current_platform
 from vllm.transformers_utils.tokenizer import get_tokenizer
@@ -99,7 +99,8 @@ class RemoteOpenAIServer:
 
         parser = FlexibleArgumentParser(
             description="vLLM's remote OpenAI server.")
-        parser = make_arg_parser(parser)
+        subparsers = parser.add_subparsers(required=False, dest="subparser")
+        parser = ServeSubcommand().subparser_init(subparsers)
         args = parser.parse_args(["--model", model, *vllm_serve_args])
         self.host = str(args.host or 'localhost')
         self.port = int(args.port)