Skip to content

Kafka


Kafka Version History

Version Release Date Key Features Introduced
0.8.0 February 2013 - Initial stable release
- Replication for fault tolerance
- New Producer API
1.0.0 November 2017 - Exactly Once Semantics (EOS)
- Performance enhancements
- Kafka Streams maturity
0.10.0.0 May 2016 - Introduction of Kafka Streams API for stream processing
1.1.0 April 2018 - Introduction of KSQL (now known as ksqlDB) for SQL-like stream processing
2.6.0 August 2020 - KIP-500 early access to remove Zookeeper dependency
- Tiered storage for cost efficiency
2.8.0 April 2021 - Continued work on KIP-500 for a simpler architecture without Zookeeper
3.0.0 September 2021 -
3.0.0 September 2021 -
3.1.0 January 2022 -
3.2.0 May 2022 -
3.3.0 28 Sep 2022 -
3.4.0 06 Feb 2023 -
3.5.0 13 Jun 2023 -
3.6.0 03 Oct 2023 -

Improvements and New Features

1. Apache Kafka 3.0.0:

Major Improvements and New Features:

  • KRaft Consensus Mechanism: Kafka 3.0 introduces improvements to KRaft, which is Kafka's built-in consensus mechanism designed to replace Apache ZooKeeper™. While KRaft is not yet recommended for production, enhancements have been made to its metadata and APIs.
  • Exactly-Once Semantics: Starting with Kafka 3.0, the producer now enables the strongest delivery guarantees by default (with acks=all and enable.idempotence=true). This ensures ordering and durability by default.
  • Partition Reassignment Support: Kafka Connect task restart enhancements, KStreams improvements in timestamp-based synchronization, and MirrorMaker2’s more flexible configuration options.

Universal Changes:

  • Deprecation of Java 8: Support for Java 8 is deprecated across all components of the Apache Kafka project in 3.0. Java 8 support is planned to be removed in the next major release (4.0).
  • Deprecation of Scala 2.12: Support for Scala 2.12 is also deprecated in Kafka 3.0, with plans to remove it in the next major release (4.0).

KRaft Snapshot:

  • KRaft controllers and brokers can now generate, replicate, and load snapshots for the metadata topic partition named __cluster_metadata. This efficient mechanism stores and replicates cluster metadata information.
    • Revised KRaft Metadata Records:
      • Improvements to metadata record types used when Kafka runs without ZooKeeper.
    • Producer ID Generation in KRaft Mode:
      • The Kafka Controller now fully handles generating producer IDs in both ZK and KRaft modes¹.

2. Apache Kafka 3.6.0:

  • Many New Features and Improvements: