Skip to main content

Architecture

jEAP Messaging organises messaging code into three layers, so that business logic stays free of Kafka boilerplate and message-type definitions can be shared across services.

Three layers​

  • jEAP layer — the jeap-messaging library itself. It provides the base message types and the pre-configured publishers and consumers. Shared across all systems.
  • System layer — the message-type definitions shared across the microservices of a system, typically in a common messages library. A message is defined as an Avro schema, usually with a generated builder and listener interface. Kafka configuration (topics, acknowledgment strategy) can also be defined here.
  • Service layer — the individual microservice. It implements message processors (consumers) and message generators (producers) as plain business logic, injecting a publisher or implementing a @KafkaListener and depending on the generated message types.

Message flow​

A producer serializes an Avro message and registers its schema in the schema registry if necessary. A consumer deserializes it. If processing fails, the error handler writes a MessageProcessingFailedEvent to the error topic instead of losing the message, and a separate Error Handling Service can resubmit it.

Module map​

The consumer-facing artifact is jeap-messaging-infrastructure-kafka, which builds on:

  • jeap-messaging-model — infrastructure-independent event/command interfaces
  • jeap-messaging-avro — Avro implementation (AvroMessage, builders, AvroSerializationHelper)
  • jeap-messaging-api — MessageListener / MessagePublisher interfaces
  • jeap-messaging-infrastructure — serialization, signing, crypto, tracing, metrics, health, error handling

Optional concerns plug in by configuration or extra dependencies: the Confluent or AWS Glue schema registry, AWS MSK IAM auth, idempotence, signing and encryption. See the full module table in jeap-messaging and Choosing dependencies.

Two registries​

Two different registries are involved, and they serve different phases:

  • The Message Type Registry is the design-time home of schemas and descriptors. It is a Git repository that validates compatibility and publishes generated Java bindings.
  • The Kafka Schema Registry (Confluent or AWS Glue) is the runtime home of schemas. jEAP Messaging registers schemas there automatically when sending; this is largely transparent to developers.