Python Alchemy缺少BEGIN似乎会导致会话不同步

更新到我的最后一个问题，请阅读背景信息。为了我的努力。我已经做了进一步的调查

TL；DR：对于某些请求，我似乎得到了旧数据。我通过发送一篇文章来更新一个对象，并有几次机会打印该对象的属性来测试这一点。通常情况下，响应是好的，但我经常会得到旧数据。当创建与其他对象有关系的对象时，这是一个大问题，如果出于某种原因外键对象不存在，创建将失败

例如：

format@pixel:~$ curl -X POST https://example.com/app/region/rename/5/Old
Success
format@pixel:~$ curl -X POST https://example.com/app/region/rename/5/New
Success
format@pixel:~$ for i in {1..25}; do  echo ; curl -X GET https://example.com/app/region/name/5/; done
New
New
New
New
New
New
Old
New
New
New
New
New
New
New
New
New
New
New
New
New
New
New
New
New
New

我在拆卸时打开了SQLALCHEMY\u ECHO和SQLALCHEMY\u COMMIT\u

app.config['SQLALCHEMY_ECHO'] = True
app.config['SQLALCHEMY_COMMIT_ON_TEARDOWN'] = True

并更改代码以打印更多信息

当请求正确时，第一个日志片段是典型的。您可以看到隐式开始和拆卸时提交

[Tue Jul 24 17:49:23.217666 2018] [wsgi:error] [pid 27919:tid 139730801542912] --- Get Start ---
[Tue Jul 24 17:49:23.219964 2018] [wsgi:error] [pid 27919:tid 139730801542912] 2018-07-24 17:49:23,219 INFO sqlalchemy.engine.base.Engine BEGIN (implicit)
[Tue Jul 24 17:49:23.220053 2018] [wsgi:error] [pid 27919:tid 139730801542912] INFO:sqlalchemy.engine.base.Engine:BEGIN (implicit)
[Tue Jul 24 17:49:23.221282 2018] [wsgi:error] [pid 27919:tid 139730801542912] 2018-07-24 17:49:23,221 INFO sqlalchemy.engine.base.Engine SELECT region.id AS region_id, region.name AS region_name, region.north AS region_north, region.west AS region_west, region.south AS region_south, region.east AS region_east, region.coords AS region_coords 
[Tue Jul 24 17:49:23.221321 2018] [wsgi:error] [pid 27919:tid 139730801542912] FROM region 
[Tue Jul 24 17:49:23.221336 2018] [wsgi:error] [pid 27919:tid 139730801542912] WHERE region.id = %s
[Tue Jul 24 17:49:23.221356 2018] [wsgi:error] [pid 27919:tid 139730801542912] 
[Tue Jul 24 17:49:23.221414 2018] [wsgi:error] [pid 27919:tid 139730801542912] INFO:sqlalchemy.engine.base.Engine:SELECT region.id AS region_id, region.name AS region_name, region.north AS region_north, region.west AS region_west, region.south AS region_south, region.east AS region_east, region.coords AS region_coords 
[Tue Jul 24 17:49:23.221433 2018] [wsgi:error] [pid 27919:tid 139730801542912] FROM region 
[Tue Jul 24 17:49:23.221447 2018] [wsgi:error] [pid 27919:tid 139730801542912] WHERE region.id = %s
[Tue Jul 24 17:49:23.221465 2018] [wsgi:error] [pid 27919:tid 139730801542912] 
[Tue Jul 24 17:49:23.221683 2018] [wsgi:error] [pid 27919:tid 139730801542912] 2018-07-24 17:49:23,221 INFO sqlalchemy.engine.base.Engine (5,)
[Tue Jul 24 17:49:23.221767 2018] [wsgi:error] [pid 27919:tid 139730801542912] INFO:sqlalchemy.engine.base.Engine:(5,)
[Tue Jul 24 17:49:23.222922 2018] [wsgi:error] [pid 27919:tid 139730801542912] 2018-07-24 17:49:23.222868 New
[Tue Jul 24 17:49:23.222990 2018] [wsgi:error] [pid 27919:tid 139730801542912] --- Get End ---
[Tue Jul 24 17:49:23.224867 2018] [wsgi:error] [pid 27919:tid 139730801542912] 2018-07-24 17:49:23,224 INFO sqlalchemy.engine.base.Engine COMMIT
[Tue Jul 24 17:49:23.224946 2018] [wsgi:error] [pid 27919:tid 139730801542912] INFO:sqlalchemy.engine.base.Engine:COMMIT

下一个代码段是请求不正确时的典型代码段。您可以看到它正在执行与以前相同的选择，但没有开始或提交。不知道这意味着什么，也不知道为什么会发生，但这让我相信问题出在SQLAlchemy或MySQL层

[Tue Jul 24 17:49:23.288116 2018] [wsgi:error] [pid 27919:tid 139730654033664] --- Get Start ---
[Tue Jul 24 17:49:23.290049 2018] [wsgi:error] [pid 27919:tid 139730654033664] 2018-07-24 17:49:23,289 INFO sqlalchemy.engine.base.Engine SELECT region.id AS region_id, region.name AS region_name, region.north AS region_north, region.west AS region_west, region.south AS region_south, region.east AS region_east, region.coords AS region_coords 
[Tue Jul 24 17:49:23.290079 2018] [wsgi:error] [pid 27919:tid 139730654033664] FROM region 
[Tue Jul 24 17:49:23.290092 2018] [wsgi:error] [pid 27919:tid 139730654033664] WHERE region.id = %s
[Tue Jul 24 17:49:23.290109 2018] [wsgi:error] [pid 27919:tid 139730654033664] 
[Tue Jul 24 17:49:23.290168 2018] [wsgi:error] [pid 27919:tid 139730654033664] INFO:sqlalchemy.engine.base.Engine:SELECT region.id AS region_id, region.name AS region_name, region.north AS region_north, region.west AS region_west, region.south AS region_south, region.east AS region_east, region.coords AS region_coords 
[Tue Jul 24 17:49:23.290186 2018] [wsgi:error] [pid 27919:tid 139730654033664] FROM region 
[Tue Jul 24 17:49:23.290197 2018] [wsgi:error] [pid 27919:tid 139730654033664] WHERE region.id = %s
[Tue Jul 24 17:49:23.290214 2018] [wsgi:error] [pid 27919:tid 139730654033664] 
[Tue Jul 24 17:49:23.290422 2018] [wsgi:error] [pid 27919:tid 139730654033664] 2018-07-24 17:49:23,290 INFO sqlalchemy.engine.base.Engine (5,)
[Tue Jul 24 17:49:23.290499 2018] [wsgi:error] [pid 27919:tid 139730654033664] INFO:sqlalchemy.engine.base.Engine:(5,)
[Tue Jul 24 17:49:23.291567 2018] [wsgi:error] [pid 27919:tid 139730654033664] 2018-07-24 17:49:23.291529 @@OLD@@
[Tue Jul 24 17:49:23.291591 2018] [wsgi:error] [pid 27919:tid 139730654033664] --- Get End ---

你可以通过时间戳看到，好的请求和坏的请求都是紧接在一起的，在这个坏的请求之后也有好几个好的请求，似乎没有韵律或理由

---

我正在使用mysqlclient 1.3.13、SQLAlchemy 1.2.10和Flask SQLAlchemy 2.3.2，并在今天早些时候运行了一个

apt-get-upgrade

如果有人发现此线程存在相同问题，请参考，我修复了我的问题

---

我的Flask App factory函数有一行

App.App\u context（）.push（）

这是早期基于Flask教程时留下的。在项目重组期间，这条线路被排除在外，问题自行解决。不确定此行为何或如何导致此问题，并且仅针对部分但并非所有请求

这对您帮助不大，但是：MySQL的默认事务隔离级别是可重复读取。因此，如果一个事务在请求之间泄漏，它将像开始时一样提供数据。提交会话自然会结束当前事务，据我所知，Flask中的“拆卸”意味着请求的结束，因此理论上您应该可以。另一方面，缺少的隐式BEGIN和COMMIT看起来好像已经有一个事务在进行中。除了前面的，Flask SQLAlchemy使用一个作用域会话，或者换句话说，一个线程本地会话注册表。因此，如果为请求提供服务的线程未删除或至少未提交/回滚会话，则下次使用该线程时，它可能正在使用现有事务。@IljaEverilä在每个请求开始时是否可以执行检查，以查看会话是否是新的？例如，对于每个连接，从文档

，会话还维护一个事务对象。

因此，如果在执行任何DB交互之前，如果OP首先检查

会话是否存在
事务，并发现存在，这将证明您的理论，对吗？Flask_SQLalchemy包含一个@app.teardown_appcontext，因此我不确定如何或为什么无法提交/回滚上一个会话，但我认为您可能走对了方向。