[Frontend] enable custom logging for the uvicorn server (OpenAI API server) (vllm-project#18403)

Signed-off-by: François Paupier <[email protected]>
Co-authored-by: Cyrus Leung <[email protected]>

Author: fpaupier

vllm/entrypoints/openai/api_server.py

@@ -5,6 +5,7 @@
 import gc
 import importlib
 import inspect
+import json
 import multiprocessing
 import os
 import signal
@@ -16,7 +17,6 @@
 from contextlib import asynccontextmanager
 from functools import partial
 from http import HTTPStatus
-from json import JSONDecodeError
 from typing import Annotated, Any, Optional
 
 import prometheus_client
@@ -930,7 +930,7 @@ async def invocations(raw_request: Request):
     """
     try:
         body = await raw_request.json()
-    except JSONDecodeError as e:
+    except json.JSONDecodeError as e:
         raise HTTPException(status_code=HTTPStatus.BAD_REQUEST.value,
                             detail=f"JSON decode error: {e}") from e
 
@@ -1003,6 +1003,18 @@ async def unload_lora_adapter(request: UnloadLoRAAdapterRequest,
         return Response(status_code=200, content=response)
 
 
+def load_log_config(log_config_file: Optional[str]) -> Optional[dict]:
+    if not log_config_file:
+        return None
+    try:
+        with open(log_config_file) as f:
+            return json.load(f)
+    except Exception as e:
+        logger.warning("Failed to load log config from file %s: error %s",
+                       log_config_file, e)
+        return None
+
+
 def build_app(args: Namespace) -> FastAPI:
     if args.disable_fastapi_docs:
         app = FastAPI(openapi_url=None,
@@ -1324,6 +1336,11 @@ async def run_server_worker(listen_address,
 
     server_index = client_config.get("client_index", 0) if client_config else 0
 
+    # Load logging config for uvicorn if specified
+    log_config = load_log_config(args.log_config_file)
+    if log_config is not None:
+        uvicorn_kwargs['log_config'] = log_config
+
     async with build_async_engine_client(args, client_config) as engine_client:
         app = build_app(args)
 

vllm/entrypoints/openai/cli_args.py

@@ -11,6 +11,7 @@
 from collections.abc import Sequence
 from typing import Optional, Union, get_args
 
+import vllm.envs as envs
 from vllm.engine.arg_utils import AsyncEngineArgs, optional_type
 from vllm.entrypoints.chat_utils import (ChatTemplateContentFormatOption,
                                          validate_chat_template)
@@ -243,6 +244,13 @@ def make_arg_parser(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
         " into OpenAI API format, the name register in this plugin can be used "
         "in ``--tool-call-parser``.")
 
+    parser.add_argument(
+        "--log-config-file",
+        type=str,
+        default=envs.VLLM_LOGGING_CONFIG_PATH,
+        help="Path to logging config JSON file for both vllm and uvicorn",
+    )
+
     parser = AsyncEngineArgs.add_cli_args(parser)
 
     parser.add_argument('--max-log-len',