There is a few more others which are more minor. So we stick with these highlights for now. If you want to know them all (also for older releases), check out the wiki: cwiki.apache.org/confluence/d...

That’s a wrap!

\TheEnd

KIP-1112: cwiki.apache.org/confluence/d...

Injecting custom boiler plate code into all Processors is now easily possible, via this new “processor wrapper” concept. Useful for PAPI users, but invaluable for the DSL!

\6

KIP-1106: cwiki.apache.org/confluence/d...

This is a shared KIP for the consumer and Kafka Streams. A new auto.offset.reset policy “by duration” is added. Very useful for tiered storage setups when you don’t want to use “earliest” but still want to start at some point in the past.

\5

KIP-1104: cwiki.apache.org/confluence/d...

Foreign-key table-table joins are very popular. However, the FK could only be extracted from the value so far. This unnecessary limitation is gone now, by passing in both key and value into the FK-key extractor.

\4

KIP-1091: cwiki.apache.org/confluence/d...

Let’s make KIP-1076 even more useful. A few key metrics like client-state and thread-state are Strings and thus cannot be collected via KIP-714. This KIP adds numerical twins for these metrics to make them collectible!

\3

KIP-1076: cwiki.apache.org/confluence/d...

This is an extension to the famous KIP-714. We now allow to register user metrics for broker side collection. Kafka Streams uses these new APIs to register its own metrics, so all Kafka Streams metrics can now be collected broker side.

\2

Note: this is the official Apache Kafka Streams account and does not endorse specific vendors, though we may share content from external sources that is relevant to open source Kafka Streams.

For hot takes and broader community updates, follow the unofficial account @thestreamingotter.bsky.social