Integration between kafka and drools

Question

I'm designing a system that needs:

• to consume events published using kafka,
• to apply on them some rules using a rule based engine
• to publish on kafka some events when the rules are matched.

To do a simple example (the real use cases are more complex): the system receives events from a temperature sensor, it uses drools to apply some rules on these events - i.e., a decision table with thresholds - and then it sends a suggestion on kafka, indicating "close the window", when the temperature goes beyond the threshold.

For what I have understood, reading the Drools documentation, currently my idea for developing this application is the following:

• Develop a Drools project inserting the rules to be applied using .drl.
• Configure Drools to work in Stream mode.
• Insert in the application also a kafka consumer, which consumes the events pubblished and then insert in drools a new event, every time it is received from kafka.
• Insert in the application a rules listener that is called every time a rule is executed.
• Publish an event on kafka every time the listener receives the events.

Is this a right way to proceed or is there another "standard" way to do this integration between kafka and drools?

Answer 1

You might consider using Quarkus, Camel and Kogito applying very similar blueprint and solution as demonstrated in:

• this blog post: https://blog.kie.org/2022/03/content-based-routing-with-quarkus-and-kogito.html
• this video: https://www.youtube.com/watch?v=nXyKM-xds2M
• this demo: https://github.com/tarilabs/quarkus-content-based-routing

The difference to compared to your use case, specifically:

To do a simple example (the real use cases are more complex): the system receives events from a temperature sensor, it uses drools to apply some rules on these events - i.e., a decision table with thresholds - and then it sends a suggestion on kafka, indicating "close the window", when the temperature goes beyond the threshold.

(emphasis mine) is that the resources above are using DMN for a decision table, but you might as well implement the same with Drools' DRL.