加固flask看门狗

Module/FlaskSubprocessManager.py

@@ -18,7 +18,7 @@ class FlaskSubprocessManager:
     """Flask 子进程守护 + 看门狗 + 稳定增强"""
 
     _instance: Optional['FlaskSubprocessManager'] = None
-    _lock = threading.Lock()
+    _lock = threading.RLock()
 
     def __new__(cls):
         with cls._lock:
@@ -42,6 +42,9 @@ class FlaskSubprocessManager:
         self._fail_count = 0
         self._last_restart_time = 0.0
 
+        self._watchdog_thread = None          # ✅ 初始化
+        self._running = False                 # ✅ 初始化
+
         # Windows 隐藏子窗口启动参数
         self._si = None
         if os.name == "nt":
@@ -175,37 +178,39 @@ class FlaskSubprocessManager:
             return False
 
     # ========= 停止 =========
-    def stop(self):
+    def stop(self, *, stop_watchdog: bool = True):
         with self._lock:
-            if not self.process: return
-            try:
-                if self.process.stdout:
-                    self.process.stdout.flush()
-                    time.sleep(0.1)  # 让读取线程跟上
-            except Exception:
-                pass
-
-        with self._lock:
-            if not self.process:
-                return
-            pid = self.process.pid
-            self._log("info", f"[FlaskMgr] 正在停止子进程 PID={pid}")
-            try:
-                parent = psutil.Process(pid)
-                for child in parent.children(recursive=True):
+            # 1) 先停子进程
+            if self.process:
+                try:
+                    self.process.terminate()
+                except Exception:
+                    pass
+                try:
+                    self.process.wait(timeout=3)
+                except Exception:
+                    pass
+                if self.process and self.process.poll() is None:
                     try:
-                        child.kill()
+                        self.process.kill()
                     except Exception:
                         pass
-                parent.kill()
-                parent.wait(timeout=3)
-            except psutil.NoSuchProcess:
-                pass
-            except Exception as e:
-                self._log("error", f"[FlaskMgr] 停止子进程异常: {e}")
-            finally:
                 self.process = None
-                self._stop_event.set()
+
+            # 2) 再考虑是否停 watchdog
+            if stop_watchdog and self._watchdog_thread and self._watchdog_thread.is_alive():
+                # 关键：不要 join 自己
+                if threading.current_thread() is not self._watchdog_thread:
+                    self._running = False
+                    try:
+                        self._watchdog_thread.join(timeout=2.0)
+                    except Exception:
+                        pass
+                    self._watchdog_thread = None
+                else:
+                    # 如果是 watchdog 自己触发的 stop，绝不 join 自己
+                    # 也不要把句柄清空，保持线程继续执行后面的重启流程
+                    self._running = True
 
     # ========= 看门狗 =========
     def _monitor(self):

Module/Main.py

@@ -30,6 +30,7 @@ print(f"日志目录: {LOG_DIR}")
 
 def _run_flask_role():
     from Module import FlaskService
+    print("Flask Pid:", os.getpid())
     port = int(os.getenv("FLASK_COMM_PORT", "34566"))  # 固定端口的兜底仍是 34567
     app_factory = getattr(FlaskService, "create_app", None)
     app = app_factory() if callable(app_factory) else FlaskService.app