Apache Kafka 3.3: What's New, Release History & Support Lifecycle

What Is New in Apache Kafka 3.3

Kafka 3.3 brings a host of improvements focused on stability, performance, and developer experience. The key changes are summarized in the table below.

Is KRaft production-ready now?

Yes, the KRaft consensus protocol is now officially production-ready in Apache Kafka 3.3. This is the headline feature, marking the culmination of the multi-year effort to remove the ZooKeeper dependency.

In practice, this means you can now run Kafka clusters without ZooKeeper for metadata management, simplifying deployment and operations. The community considers it stable for production workloads, though many will take a cautious rollout approach.

What tools help with the ZooKeeper migration?

A new CLI tool, kafka-metadata-shell, was introduced to help inspect and debug the metadata state in KRaft clusters. This is crucial for operators transitioning from ZooKeeper.

You can use it to directly examine the metadata log, query partition states, and understand the internal snapshot files. It fills a gap that existed during the KRaft development cycle, providing much-needed visibility.

How is partition management improved?

The partition reassignment process received significant enhancements. The kafka-reassign-partitions.sh tool now supports a --verify option and provides more detailed progress output.

This matters because reassignments are a common but potentially risky operational task. The improved feedback loop makes it easier to track the movement of data and confirm completion without guessing.

Were there any performance tweaks?

Yes, several under-the-hood changes improve performance and stability. The default behavior for log flushing was adjusted to be more efficient, which can reduce I/O pressure.

There are also updates to how ZooKeeper session timeouts are handled, reducing unnecessary churn. These are the kinds of incremental gains that add up to a more robust cluster in high-throughput scenarios.