How to solve the data inconsistency problem: an external call was successful, but saving to the database failed?

Data inconsistency issues, where an external call succeeded but saving to the database failed, are addressed using transaction management patterns and compensating mechanisms.

Main approaches:

1. Transactions and rollback: If the external call and database save need to be atomic, distributed transactions (such as XA transactions or two-phase commit) should be used. However, this is complex and not always feasible.

2. Compensating operations (SAGA): Break the process into steps, each of which can be compensated. If saving to the database fails, invoke a compensating call to cancel the external call.

3. Idempotency and retries: Make operations idempotent so that repeated calls do not cause side effects.

4. Queues and asynchrony: Instead of directly calling an external service, send an event to a queue, and handle and save in a single transactional process.

Example with SAGA (simplified):

// External service call
boolean externalSuccess = externalService.call();
if (externalSuccess) {
    try {
        database.save(data);
    } catch (Exception e) {
        // Compensation: cancel external call
        externalService.compensate();
        throw e;
    }
}

Thus, it is important to provide a compensation mechanism or ensure the atomicity of operations.