I have a question about what happens if I deploy a postgres cluster following the recommended approach: Multi-region PostgreSQL · Fly
Does fly ensure that write requests don’t complete until after it’s been propagated to every region? If not, how do we avoid inconsistencies? If a secondary region makes a write to the primary region and then immediately does a read of its closer region before the change has been propagated then it would get old data.
So my question is, does fly ensure writes don’t complete until propagation is completed? If not, how are inconsistencies avoided?