How were deviations handled when working with Kafka and REST (message did not arrive, service did not respond)?

When working with Kafka and REST for handling failures (for example, a message not received or a service not responding), the following approaches are used:

• For Kafka:

  • Using acknowledgments (acks) from the broker to guarantee delivery.
  • Implementing retries for message sending upon errors.
  • Logging and monitoring errors.
  • Using the "Dead Letter Queue" (DLQ) scheme for messages that could not be delivered after several attempts.

• For REST:

  • Timeouts and retries for requests upon failure.
  • Handling HTTP status errors (e.g., 5xx — server errors, 4xx — client errors).
  • Using the Circuit Breaker scheme to prevent service overload.
  • Logging and alerting on errors.

Example: if an error occurs when sending a message to Kafka, the service can retry several times with exponential backoff. If all attempts fail, the message is placed in the DLQ for later analysis. Similarly, when calling a REST API, if the service does not respond, the client retries the request a limited number of times and logs the error for further investigation.