How do transactions work in Spring?

A transaction provides atomicity, consistency, isolation and durability. Spring's @Transactional starts a transaction before the method and commits or rolls back after it. The annotation works through a proxy, so calls that bypass the proxy do not get transactional behaviour.

@Service
public class TransferService {
    @Transactional
    public void transfer(long fromId, long toId, long amount) {
        Account from = accounts.findById(fromId).orElseThrow();
        Account to = accounts.findById(toId).orElseThrow();
        from.withdraw(amount);
        to.deposit(amount);
    }
}

Settings that change behaviour

• Propagation: REQUIRED joins an existing transaction; REQUIRES_NEW suspends it
• Isolation: READ_COMMITTED is common; higher levels reduce anomalies but increase contention
• Rollback: unchecked exceptions roll back by default, checked exceptions do not unless configured
• readOnly is a hint that can skip dirty checking and route to a replica
• Self-invocation bypasses the proxy, so an internal call to a transactional method is not transactional

Keep transactions short and avoid remote calls inside them, because a slow dependency holds locks and connections. Put the transactional boundary in the service layer, not the controller, and be explicit about rollback rules when checked exceptions are part of the contract.