SQLAlchemy并发和查询

Question

I have a number of separate machines, each of which uses a separate instantiation of

Session = scoped_session(sessionmaker(bind=engine)) .

If machine A runs the code

session = Session()
session.add(Foo(pk=1))
session.commit()

what do I have to do ensure that when machine B runs the below code...

session = Session()
result = session.query(Foo).get(1)

... that result holds the new row of Foo that machine A created (assume that machine A's code ran first). I've had problems before where Session "A" commits an object but another Session "B" cannot find it in a query until I reinstantiate Session "B".

Answer 1

scoped_session wraps Session in a thread local storage object; if you're doing more work to handle threads, you might want to either refactor your code to use a single scoped_session instance for all threads, or use the result of sessionmaker directly, without wrapping it, again one for all threads, and call Session() for each thread.