At Klarna, Lead Engineer Tommy Brunn is building a runtime platform for developers. But outside of his professional role, he is also one of the authors of the JavaScript client for Apache Kafka® called KafkaJS, which has grown from being a niche open source project to the most downloaded Kafka client for Node.js since 2018.
Using Kafka in Node.js has previously meant relying on community-contributed bindings to librdkafka, which required you to spend more of your time debugging failed builds than working on your application. With the original authors moving away from supporting the bindings, and the community only partially picking up the slack, using Kafka on NodeJS was a painful proposition.
Kafka is a core part of Klarna’s microservice architecture, with hundreds of services using it to communicate among themselves. In 2017, as their engineering team was building the ecosystem of Node.js services powering the Klarna app, it was clear that the experience of working with any of the available Kafka clients was not good enough, so they decided to perform something similar for the Erlang client, Brod, and build their own. Rather than wrapping librdkafka, their client is a complete reimplementation in native JavaScript, allowing for a far superior user experience at the cost of being a lot more work to implement. Towards the end of 2017, KafkaJS 0.1.0 was released.
Tommy has also used KafkaJS to build several Kafka-powered services at Klarna, as well as worked on supporting libraries such as integrations with Confluent Schema Registry and Zstandard compression.
Since KafkaJS is written entirely in JavaScript, there is no build step required. It will work 100% of the time in any version of Node.js and evolve together with the platform with no effort required from the end user. It also unlocks some creative use cases. For example, Klarna once did an experiment where they got it to run in a browser. KafkaJS will also run on any platform that’s supported by Node.js, such as ARM. Klarna’s “no dependencies” policy also means that the deployment footprint is small, which makes it a perfect fit for serverless environments.
EPISODE LINKS