apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-12-24 13:28:13 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

Improve error handling in cleanup functions

Browse Source 
- Add check for already terminated processes before attempting cleanup
- Add ProcessLookupError handling for race conditions
- Add success logging for better debugging
- Check if cache manager processes are alive before terminating

Co-authored-by: Jiang-Jia-Jun <163579578+Jiang-Jia-Jun@users.noreply.github.com>
This commit is contained in:
copilot-swe-agent[bot]
2025-12-23 12:13:11 +00:00
parent b68b7c8688
commit cd62cc2df9
2 changed files with 23 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
fastdeploy/entrypoints/api_server.py
View File

@@ -43,9 +43,19 @@ def cleanup_engine():
try:
if hasattr(llm_engine, "worker_proc") and llm_engine.worker_proc is not None:
try:
# 检查进程是否已经结束
if llm_engine.worker_proc.poll() is not None:
api_server_logger.info("Worker process already terminated")
return
pgid = os.getpgid(llm_engine.worker_proc.pid)
api_server_logger.info(f"Terminating worker process group {pgid}")
os.killpg(pgid, signal.SIGTERM)
# 等待进程结束
llm_engine.worker_proc.wait(timeout=5)
api_server_logger.info("Worker process terminated successfully")
except ProcessLookupError:
api_server_logger.info("Worker process already terminated")
except Exception as e:
api_server_logger.error(f"Error terminating worker process: {e}")
except Exception as e:

16
fastdeploy/entrypoints/openai/api_server.py
View File

@@ -97,26 +97,36 @@ def cleanup_processes():
try:
if hasattr(llm_engine, "worker_proc") and llm_engine.worker_proc is not None:
try:
# 检查进程是否已经结束
if llm_engine.worker_proc.poll() is not None:
api_server_logger.info("Worker process already terminated")
return
pgid = os.getpgid(llm_engine.worker_proc.pid)
api_server_logger.info(f"Terminating worker process group {pgid}")
os.killpg(pgid, signal.SIGTERM)
# 等待进程结束，如果超时则强制杀死
try:
llm_engine.worker_proc.wait(timeout=5)
api_server_logger.info("Worker process terminated successfully")
except subprocess.TimeoutExpired:
api_server_logger.warning("Worker process did not terminate in time, sending SIGKILL")
os.killpg(pgid, signal.SIGKILL)
llm_engine.worker_proc.wait(timeout=2)
except ProcessLookupError:
api_server_logger.info("Worker process already terminated")
except Exception as e:
api_server_logger.error(f"Error terminating worker process: {e}")
if hasattr(llm_engine, "cache_manager_processes") and llm_engine.cache_manager_processes:
for proc in llm_engine.cache_manager_processes:
try:
proc.terminate()
proc.join(timeout=2)
if proc.is_alive():
proc.kill()
proc.terminate()
proc.join(timeout=2)
if proc.is_alive():
proc.kill()
proc.join(timeout=1)
except Exception as e:
api_server_logger.error(f"Error terminating cache manager process: {e}")
except Exception as e: