[Feat] Allow custom naming of vLLM processes

vllm/entrypoints/openai/api_server.py

@@ -9,6 +9,7 @@
 import json
 import multiprocessing
 import os
+import setproctitle
 import signal
 import socket
 import tempfile
@@ -1803,7 +1804,8 @@ async def run_server_worker(listen_address,
         ToolParserManager.import_tool_parser(args.tool_parser_plugin)
 
     server_index = client_config.get("client_index", 0) if client_config else 0
-
+    setproctitle.setproctitle(
+        f"{envs.VLLM_PROCESS_NAME_PREFIX}::APIServer_{server_index}")
     # Load logging config for uvicorn if specified
     log_config = load_log_config(args.log_config_file)
     if log_config is not None:

vllm/envs.py

@@ -971,6 +971,12 @@ def get_vllm_port() -> Optional[int]:
     # Used to force set up loopback IP
     "VLLM_LOOPBACK_IP":
     lambda: os.getenv("VLLM_LOOPBACK_IP", ""),
+
+    # Used to set the process name prefix for vLLM processes.
+    # This is useful for debugging and monitoring purposes.
+    # The default value is "VLLM".
+    "VLLM_PROCESS_NAME_PREFIX":
+    lambda: os.getenv("VLLM_PROCESS_NAME_PREFIX", "VLLM"),
 }
 
 # --8<-- [end:env-vars-definition]

vllm/v1/engine/core.py

@@ -6,6 +6,7 @@
 import sys
 import threading
 import time
+import setproctitle
 from collections import deque
 from collections.abc import Generator
 from concurrent.futures import Future
@@ -17,6 +18,7 @@
 import msgspec
 import zmq
 
+import vllm.envs as envs
 from vllm.config import ParallelConfig, VllmConfig
 from vllm.distributed import stateless_destroy_torch_distributed_process_group
 from vllm.executor.multiproc_worker_utils import _add_prefix
@@ -572,7 +574,8 @@ def run_engine_core(*args,
                         local_dp_rank: int = 0,
                         **kwargs):
         """Launch EngineCore busy loop in background process."""
-
+        setproctitle.setproctitle(
+            f"{envs.VLLM_PROCESS_NAME_PREFIX}::EngineCore_{dp_rank}")
         # Signal handler used for graceful termination.
         # SystemExit exception is only raised once to allow this and worker
         # processes to terminate without error

vllm/v1/executor/multiproc_executor.py

@@ -9,6 +9,7 @@
 import time
 import traceback
 import weakref
+import setproctitle
 from concurrent.futures import Future, ThreadPoolExecutor
 from dataclasses import dataclass
 from enum import Enum, auto
@@ -365,7 +366,9 @@ def __init__(
         }
         wrapper.init_worker(all_kwargs)
         self.worker = wrapper
-
+        setproctitle.setproctitle(
+            f"{envs.VLLM_PROCESS_NAME_PREFIX}::{
+                self.worker.worker.__class__.__name__}_{self.rank}")
         pid = os.getpid()
         _add_prefix(sys.stdout, f"VllmWorker rank={rank}", pid)
         _add_prefix(sys.stderr, f"VllmWorker rank={rank}", pid)