In this 21st edition of the Kafka Monthly Digest, I will cover what happened in the Kafka Community in October 2019.
For last month’s digest, see “Kafka Monthly Digest: September 2019.”
Releases
- 2.3.1: This bugfix version was released on October 25, 2019 by David Arthur. It contains 46 fixes and improvements, including six blocker JIRAs. See the release notes for the full details.
There are still one bugfix release and one minor release in progress.
- 2.2.2: After a couple of late bugs (KAFKA-9053, KAFKA-8649) related to Kafka Streams had been fixed, Randall Hauch prepared 2.2.2 RC2. The vote on this release candidate is currently ongoing. If no further issues are found, it should release in the next few days.
- 2.4.0: The release process continued, and the code freeze happened on October 16, 2019. A few issues were found and there is currently one blocker left open (KAFKA-8677). The first release candidate should be available in the next few days.
KIPs
Last month, the community submitted 17 KIPs (KIP-530 to KIP-546, 529 was skipped) and these are the ones that caught my eye:
KIP-531: Drop support for Scala 2.11 in Kafka 2.5: At the moment, Kafka builds for three Scala versions: 2.11, 2.12, and 2.13. The proposal is to stop building for 2.11 as this version has reached end of support. This will simplify the compatibility matrix and reduce development and tests costs. Finally, this will allow for taking advantage of more recent Scala features, especially around interoperability with Java 8 functional interfaces.
KIP-542: Partition Reassignment Throttling: Partition reassignments often cause significant load on a cluster. Since 0.10.1, quotas can be applied to control them, but these quotas also apply to existing replicas. The goal of this KIP is to allow for defining separate quotas for replication and reassignment. This will enable administrators to control reassignments without impacting regular replication.
KIP-543: Expand ConfigCommand’s non-ZK functionality: One of the pre-requirements to remove Zookeeper (KIP-500) is to update all tools to work without Zookeeper. At the moment, the kafka-configs tool still requires Zookeeper to update topic configurations and quotas. This KIP’s goal is to update this tool so it does not require on Zookeeper for any operations.
KIP-544: Make metrics exposed via JMX configurable: Kafka exposes a lot of metrics. In fact, in large clusters, monitoring agents can sometimes timeout while listing metrics. This KIP proposes adding configurations to the broker to blacklist or whitelist metrics. By also allowing to update these settings at runtime, administrators will be able to specify the metrics they want to omit, precisely.
KIP-545: support automated consumer offset sync across clusters in MM 2.0: MirrorMaker 2 added some tooling to translate consumer group offsets between clusters. This KIP aims to make failover between clusters even simpler by automatically populating the __consumer_offsets topics in downstream clusters. This will allow consumers to transparently find their new positions when switching clusters.
Blogs
- Kafka Summit 2019 San Francisco: A review
- How LinkedIn customizes Apache Kafka for 7 trillion messages per day
- Why I Recommend My Clients NOT Use KSQL and Kafka Streams
- Event-Driven Architecture and the Outbox Pattern
Get started
IBM Event Streams for Cloud is Apache Kafka-as-a-service for IBM Cloud.