Get Started Free

Correlation Identifier

Event Processing Applications may want to implement Event Collaboration, where Events are used to transport requests and responses. When applications collaborate via events, they need a way to correlate event response data for specific requests.

Problem

When an application requests information and receives a response, how can the application know which request corresponds to that particular response?

Solution

correlation-identifier

An Event Processor generates an event, which acts as the request. A globally unique identifier is added to the request event before it is sent. This allows the responding Event Processor to include the identifier in the response event, so that the requesting processor can correlate the request and response.

Implementation

In Kafka, we can add a globally unique identifier to Kafka record headers when producing a request event. The following example uses the Kafka Java producer client.

ProducerRecord<String, String> requestEvent = new ProducerRecord<>("request-event-key", "request-event-value"); 
requestEvent.headers().add("requestID", UUID.randomUUID().toString());
requestEvent.send(producerRecord);

In the responding event processor, we first extract the correlation identifier from the request event (here, the identifier is called requestID) and then add that identifier to the response event.

ProducerRecord<String, String> responseEvent = new ProducerRecord<>("response-event-key", "response-event-value"); 
responseEvent.headers().add("requestID", requestEvent.headers().lastHeader("requestID").value());
responseEvent.send(producerRecord);

References

  • This pattern is derived from Correlation Identifier in Enterprise Integration Patterns, by Gregor Hohpe and Bobby Woolf.
  • For a case study on coordinating microservices towards higher-level business goals, see Building a Microservices Ecosystem with Kafka Streams and ksqlDB.
  • Correlation identifiers can be used as part of Event Collaboration, a pattern in which decentralized Event Processing Applications collaborate to implement a distributed workflow solution.
  • The idea of tagging requests and their associated responses exists in many other protocols. For example, an email client connecting over IMAP will send commands prefixed with a unique ID (typically a001, a002, etc.), and the server will respond asynchronously, tagging its responses with the matching ID.

Confluent Cloud is a fully managed Apache Kafka service available on all three major clouds. Try it for free today.

Try it for free