Distributed Transaction
A logical operation whose correctness spans multiple independently failing databases, resource managers or services.
Problem Boundary
A distributed transaction appears when one business operation requires a coordinated outcome across multiple independently failing participants. Unlike a local database transaction, network delay, partial failure and temporary unavailability are part of the correctness model.
Coordination Approaches
Two-Phase Commit targets atomic commit across participants. A Saga decomposes the workflow into local transactions with compensating actions. Transactional Outbox solves a different reliability problem by committing business state and an outbound-message record locally before asynchronous publication.
These approaches are not interchangeable; they provide different semantics for atomicity, availability, resource retention and recovery.
Common Misconception
Calling two services sequentially over HTTP does not automatically create a transaction. The system has to define what happens when only some steps complete and which effects can be retried, reversed or compensated.