How was the interaction divided — what went through Kafka, and what was direct?

In distributed systems, interaction via Kafka is typically used for asynchronous message exchange, where reliability, scalability, and real-time event processing are important. For example, events that need to be processed by multiple services or require guaranteed delivery are sent through Kafka.

Direct communication (e.g., via REST API or gRPC) occurs when a synchronous request-response is needed, such as retrieving data upon a user request or calling an operation that requires an immediate result.

Thus:

• Kafka is used for events, logs, asynchronous notifications, and background data processing.
• Direct interaction is used for synchronous requests when a quick response or transactional logic is required.

This separation helps to unload services, increase fault tolerance, and improve system scalability.