How to Effectively Handle Network Exceptions in an MSA Environment

In a microservices architecture (MSA), each service manages an independent DB, so ensuring data consistency for global transactions spanning multiple services is very important. Especially in the payment domain, network latency or timeouts can cause situations where it is unclear whether a request succeeded.

The result of a network request is not simply divided into success and failure; an 'Unknown' state, where the result is unclear, must also be considered. To address this, the requesting side corrects the transaction through post-processing such as immediate retry, compensating transactions, or manual handling.

From the API-provider side, securing Idempotency is key. By designing the system so that the same request returns the same response even when it comes in multiple times, problems such as duplicate payments caused by retries can be prevented. To achieve this, it is important to include a unique key (tx_key) in API requests to identify whether the request is the same.