fix(atest): 去除 manage.py 硬编码密钥口令 + SIO 初次连接失败兜底重连

- manage.py 强制环境变量注入，删除默认 SECRET_KEY/MSYNC_SERVICE_SECRET/DB_PASSWORD
- sio_client 初次连接失败启动兜底重连线程，消除永久退化态
- 补充 RetryStep 手动重试与状态守卫回归用例（17/17 通过）

ATestCServer/manage.py

@@ -7,9 +7,16 @@ def main():
     sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
     os.environ['DJANGO_SETTINGS_MODULE'] = 'ATestCServer.settings'
     os.environ['DB_NAME'] = 'atestc'
-    os.environ.setdefault('DJANGO_SECRET_KEY', 'gvsds-django-prod-k8x2m9v4n7p1q6w3j5r0t8y1u4i7o0a2s6d9f3g5h7j2k4l6z8x1c3v5b')
-    os.environ.setdefault('MSYNC_SERVICE_SECRET', 'atestc-msync-secret')
-    os.environ.setdefault('DB_PASSWORD', '1Qq@1Qq@190*')
+    # 安全：禁止硬编码默认密钥/口令，缺失时显式报错（settings.py 亦有
+    # ImproperlyConfigured 兜底）。MSYNC_SERVICE_SECRET 与 DB_PASSWORD
+    # 由部署环境（环境变量）提供。
+    required_envs = ['DJANGO_SECRET_KEY', 'MSYNC_SERVICE_SECRET', 'DB_PASSWORD']
+    missing = [name for name in required_envs if not os.environ.get(name)]
+    if missing:
+        raise SystemExit(
+            '缺少必需的环境变量: ' + ', '.join(missing)
+            + '（禁止使用默认值，请通过环境变量注入）'
+        )
     try:
         from django.core.management import execute_from_command_line
     except ImportError as exc:

ATestCServer/msync_client/sio_client.py

@@ -103,10 +103,40 @@ class MSyncSocketClient:
             logger.info(f'[SIO] → connect url={url} ns=/msync')
             self._sio.connect(url, namespaces=['/msync'], wait=True, wait_timeout=15)
         except Exception as e:
-            logger.warning(f'Socket.IO 连接失败: {e}，将自动重连')
-            self._StartHeartbeat()
+            logger.warning(f'Socket.IO 连接失败: {e}，启动兜底重连线程')
+            self._StartBootstrap()
             return
 
+        self._OnReady()
+
+    def _StartBootstrap(self):
+        """初次连接失败的兜底：循环重连，成功后走完整注册/认证流程
+
+        python-socketio 的 reconnection 仅在"已连接后断开"时生效，
+        初次 connect 失败不会自动重连，需自行兜底。
+        """
+
+        def bootstrap():
+            while self._running and not self._sio.connected:
+                try:
+                    logger.info(f'[SIO] 兜底重连尝试: {self.MasterURL}')
+                    self._sio.connect(
+                        self.MasterURL,
+                        namespaces=['/msync'],
+                        wait=True,
+                        wait_timeout=10,
+                    )
+                    break
+                except Exception as e:
+                    logger.warning(f'[SIO] 兜底重连失败: {e}')
+                    time.sleep(3)
+            if self._sio.connected:
+                self._OnReady()
+
+        threading.Thread(target=bootstrap, daemon=True, name='msync-bootstrap').start()
+
+    def _OnReady(self):
+        """连接建立后的统一就绪流程：注册/认证/心跳/ORM 中转"""
         logger.info(f'[SIO] Socket 已连接, 开始注册/认证...')
         for attempt in range(1, 4):
             reg_ok = self._DoRegister()

gvsds/tests/test_msync_saga.py

@@ -53,7 +53,7 @@ def _Run():
 
     from unittest import mock
     import uuid
-    from gvsdsdk.msync.transaction import TransactionManager, StepExecutionError
+    from gvsdsdk.msync.transaction import TransactionManager, StepExecutionError, TransactionError
     from gvsdsdk.msync.local_tx import local_tx
     from gvsdsdk.msync.models import ServiceNode
 
@@ -164,6 +164,53 @@ def _Run():
     r3 = local_tx.ExecuteStep('no_such', {}, 'K:1:1')
     check('不同键不命中', r1 is not r3)
 
+    # ─── 6. RetryStep 手动重试（自动重试耗尽后仍可重试，计数分离）───
+    print('[6] RetryStep 手动重试')
+    dtx = TransactionManager.Create(
+        name='t6',
+        steps_data=[
+            {'name': 's1', 'ParticipantService': 'svc'},
+            {'name': 's2', 'ParticipantService': 'svc', 'MaxRetries': 0},
+        ],
+    )
+    with mock.patch.object(TransactionManager, '_call_service', side_effect=[
+        {'ok': 1},
+        StepExecutionError('s2 永久失败'),       # s2 自动重试耗尽
+        StepExecutionError('补偿 s1 也失败'),    # 补偿失败 → 事务 FAILED
+    ]):
+        TransactionManager.Execute(dtx.TransactionCode)
+    dtx.refresh_from_db()
+    check('自动重试耗尽后事务 Failed', dtx.TransactionStatus == 5)
+    # s2 MaxRetries=0 已耗尽，原逻辑（CanRetry）拒绝；现手动重试不受配额限制
+    with mock.patch.object(TransactionManager, '_call_service', return_value={'ok': True}):
+        step = TransactionManager.RetryStep(dtx.TransactionCode, 1)
+    check('手动重试成功 Completed', step.StepStatus == 2)
+    check('手动重试计数独立', step.RetryCount == 1)
+
+    # ─── 7. RetryStep 拒绝已补偿事务（事务状态守卫）───
+    print('[7] RetryStep 已补偿事务拒绝')
+    dtx = TransactionManager.Create(
+        name='t7',
+        steps_data=[
+            {'name': 's1', 'ParticipantService': 'svc'},
+            {'name': 's2', 'ParticipantService': 'svc', 'MaxRetries': 0},
+        ],
+    )
+    with mock.patch.object(TransactionManager, '_call_service', side_effect=[
+        {'ok': 1},
+        StepExecutionError('s2 失败'),
+        {'ok': True},  # s1 补偿成功 → 事务 COMPENSATED
+    ]):
+        TransactionManager.Execute(dtx.TransactionCode)
+    dtx.refresh_from_db()
+    check('事务已补偿', dtx.TransactionStatus == 4)
+    rejected = False
+    try:
+        TransactionManager.RetryStep(dtx.TransactionCode, 1)  # s2 正向失败步骤
+    except TransactionError:
+        rejected = True
+    check('已补偿事务 RetryStep 被拒绝', rejected)
+
     # ─── 汇总 ───
     passed = sum(1 for _, ok in results if ok)
     print(f'\n结果: {passed}/{len(results)} 通过')