[Bugfix] use AF_INET6 instead of AF_INET for OpenAI Compatible Server

vllm/entrypoints/openai/api_server.py

@@ -11,7 +11,7 @@
 from contextlib import asynccontextmanager
 from functools import partial
 from http import HTTPStatus
-from typing import AsyncIterator, Set
+from typing import AsyncIterator, Set, Tuple
 
 import uvloop
 from fastapi import APIRouter, FastAPI, Request
@@ -524,6 +524,17 @@ def init_app_state(
     )
 
 
+def create_server_socket(addr: Tuple[str, int]) -> socket.socket:
+    if socket.has_dualstack_ipv6():
+        sock = socket.create_server(addr,
+                                    family=socket.AF_INET6,
+                                    dualstack_ipv6=True)
+    else:
+        sock = socket.create_server(addr)
+
+    return sock
+
+
 async def run_server(args, **uvicorn_kwargs) -> None:
     logger.info("vLLM API server version %s", VLLM_VERSION)
     logger.info("args: %s", args)
@@ -540,8 +551,8 @@ async def run_server(args, **uvicorn_kwargs) -> None:
     # workaround to make sure that we bind the port before the engine is set up.
     # This avoids race conditions with ray.
     # see https://github.com/vllm-project/vllm/issues/8204
-    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
-    sock.bind(("", args.port))
+    sock_addr = ("", args.port)
+    sock = create_server_socket(sock_addr)
 
     def signal_handler(*_) -> None:
         # Interrupt server on sigterm while initializing