一次停止运行一次以上的Django Model函数

Question

I have a Django Model class for a booking slot. Once we have all the auxiliary data and payment is ready to be made, we get a live availability and allocate a booking. This whole last step of the process takes a couple of seconds (mostly waiting on the payment provider to clear). In theory two payments coming through at once could double-book a slot.

That's all handled by a single function Booking.book() . Is there any sane way I can limit so that only one instance can work at once and others are queued?

The deployment design is initially pretty simple but there could be scale to multiple servers eventually.

What's the proper way of doing this and what are its downsides?

Answer 1

What you can do is lock the slot for the booking once the payment is initiated. That way you only "lose" availability for a few moments. The lock can be done on the same centralized system that holds the rest of the information. For instance you can scale up the application servers but keep a single entry point to the data source. You release the lock once the payment is declined or confirm.

Answer 2

You could add a foreignkey to your products like this :

class Reservation(models.Model):
    reservations = ForeignKey(Book)
    quantity = models.IntegerField(default=0)
    creation = models.DateTimeField(auto_now_add)
    user = ForeignKey(User)

class Book(models.Model):
    ...

Whenever someone is ready for a payment, you add a Reservation per book type, and reduce your Book.stock by the quantity (for this part, you could also have a "reserved" column if you prefer, or calculate the available quantity using joins but this last would'nt be very efficient).

If the payment is validated, you only have to remove the Reservation entries, and if it is not, put back your stock. You could have a cron every few minutes that would get the stocks back if the reservation is expired (using the creation field).