Since 4.1.0 Kafka we have an opportunity to test totally new approach for consuming messages from Kafka:

We can consume data from the single topic partition by multiple Kafka Consumer instances running in the same Consume Group!

This is a huge change in the matter of dynamic scaling Kafka Consumers, before 4.1.0 maximum level of parallel processing was limited by number on topics partitions we are listening on, with that release we try to scale our applications dynamically, based on some different factors like number of messages or Consumer Group lag.

You can read more about Queues in Apache Kafka in official Kafka Improvement Proposal documents:

1) KIP-932: Queues for Kafka

2) Queues for Kafka (KIP-932) - Early Access Release Notes

So lets play with Shared Kafka Consumer:

1. Download Apache Kafka binaries https://kafka.apache.org/downl...
2. Modify /config/server.properties file, add following configurations:
   unstable.api.versions.enable=true
   group.coordinator.rebalance.protocols=classic,consumer,share
   group.share.enable=true
3. Run Kafka Broker
4. Create topic
   1. .bin/kafka-topics.sh --bootstrap-server localhost:9092 --create --topic my-queue
5. Clone github repository https://github.com/pszymczyk/k...
6. Open project in your favourite IDE
7. Navigate to com.pszymczyk.step6.Step6 class
8. Navigate to line 50 and see that we are using org.apache.kafka.clients.consumer.ShareConsumer<String, String> consumer instead of classic Kafka Consumer 
9. Navigate to line 60 and see that we provided new Shared Kafka Consumer configuration parameter props.put(SHARE_ACKNOWLEDGEMENT_MODE_CONFIG, "explicit") so we can use totally new Shared Consumer API consumer.acknowledge(record, AcknowledgeType.ACCEPT) (line 78).
10. Run SubscribeRunner.main()
11. Analyze logs, you can see that different members are assigned to the same topic partition

12. Check consumer group state and members using Kafka CLI, you should also see that we have 3 members assigned to the same topic and partition:

13. Finally produce some messages by running com.pszymczyk.step6.PublishRunner. In the application logs you will see that our application is consuming messages from single topic partition on multiple Kafka Consumer instances running on different threads (not possible before 4.1.0 release)