I am working at a Spring Java EE application that has to deal with two different datasources, A and B.
As the application needs to update A and B in a consistent way. If one update fails, the whole process fails and a rollback must take place.
I have two different idea on how to implement the application:
What do you think about those two scenario?
A query that uses a DB link will start a distributed transaction in Oracle implicitly. So you get the same overhead as dealing with XA and two datasources from your application.
We are using XA all the time, and it has never been a bottleneck. Don't worry.