Normally you don't setup services in the processes as they are created automatically based on the (supported) framework used. You can use Custom services or Oneagent SDK to setup services that are were not discovered or were not discovered as you need them to be.
Were at least some services created for your Kafka and Zookeeper processes? Kafka or Zookeper are not within officially supported technologies (only JMX metrics for Kafka). Afaik Dynatrace Oneagent instruments some classes and methods for Kafka, I have not been successful in getting (autodetected) services for Kafka consumers or producers in my lab.
In general Kafka and Zookeeper are monitored not via services. But there are informations from CPU sampling (method hotspots), visible from process card.
What I’ve done on our client was instrumenting peace’s of code that are sending requests to Kafka from application endpoints. There I see performance and responsiveness of Kafka.
For Kafka JMX measures + infrastructure are good enough for troubleshooting.