Will KIP-974 obviate the need for this project?

[k-wall]

I've really appreciated the trail blazing this project has done in helping make native Kafka a thing in the community.

Now that KIP-974 has materialised and Apache Kafka publishes its own native images, how much value does this project continue to have? Should it sunset?

I can think of some use-cases that this project provides that aren't covered by KIP-974.

• this project provides a simplified interface for starting a broker.
• this project includes some Strimzi SASL plugins that aren't included in Apache Kafka native.
• this project provides a native Zookeeper which is valuable for those testing pre-KRaft use-cases.

Are there others?

[ozangunalp]

I agree with this. I've been trying to find the time to look into this. I can add more into the pro-list:

• Build-time init approach brought by Quarkus allows consuming less memory at startup. When compared the RSS of this image is 3 times less than the official kafka-native image. It would be interesting to bench this more and contribute some build time init back to upstream Kafka native build.

As for the cons :

• Maintenance burden, with the coming 4.0 release
• Image composition doesn't follow standard Apache Kafka distributions. For example, people have been looking for some scripts to run kafka command line tools are not a part of the docker image #192, Docker compose health checks #216

[k-wall]

Yes, it would be fantastic to see those RSS improvements back in upstream kafka's native image. I wish I could say I could help, but it is not really part of my skill set. If there's a break down of tasks, I might see if I can help with some of the lifting.

I agree with the cons you've highlighted. Having an Kafka image with a well known command line interface so really going to help the community innovate. So I think long term having ozangunalp/kafka-native out there too with a different API probably isn't too helpful.

That said, it would really help the Kroxylicious project if this project could produce a Kafka 3.9 image (when 3.9 comes out). We'll probably use the point Kafka 4.0 is released as the trigger to move to testing with Apache Kafka images.

[ozangunalp]

That said, it would really help the Kroxylicious project if this project could produce a Kafka 3.9 image (when 3.9 comes out). We'll probably use the point Kafka 4.0 is released as the trigger to move to testing with Apache Kafka images.

That's a nice plan.