Tom Cooper

YouTube video by Strimzi Phased upgrades of Strimzi managed Kafka fleets

A bit late, but the recording of my #StrimziCon2025 talk, Phased Upgrades of Strimzi Managed Kafka Fleets (hows that for a snappy title) is now up on Youtube: youtu.be/pEuYFFkQCUU?...

It's my last day at Red Hat. Tomorrow I & the rest of middleware move to IBM. I wanted to work at RH since I heard about Fedora on the Linux Action Show (RIP) decades ago & I've loved developing in the open with RH. That'll continue after the move & after all, no one got fired for buying IBM...

🐊 Kroxylicious Proposal: "Routing API"

This looks like a very useful addition to @kroxylicious.io (wire-protocol proxy for Apache Kafka): dynamic routing of requests to different upstream clusters, enabling use cases such as union clusters and client-based routing.

👉 github.com/kroxylicious...

Going from 1 to 13 PB/day 🤯 thanks to moving from a self-managed to a Strimzi-managed Kafka cluster! Great to hear these stories of how Strimzi not only made life easier for infra teams but improved the service they provide!

strimzi.io/blog/2025/06...

It was a really great talk, you covered a lot in 15mins! I am keen to work on an OSS library for embeddings wrapping Langchain4j so may ping you for opinions/reviews if you have time?

Luckily we have 2 people on the Strimzi team who literally wrote the book on Kafka Connect and Strimiz could do something (TBD) to help.

Mostly about how everyone wants DR, try MM2, bounce off it and then end up with some very unique setups and trade offs. TL;DR MM2 (and connect more generally) is hard to get started with and has some sharp edges.

"Process" not "Polymorphic" 🤦‍♂️

Ending #Current25 with a talk on Polymorphic Table Functions in Flink🤘

Great hallway track at #Current25, lots of ideas around Kafka connect, mirror maker, geo-replication, cluster balancing. Lots to mull over on the train home...

"Vibe-coded database schemas" 😱

So nice to get a deep technical deep dive on a subject. They did a great job of explaining the migration process. If you want more detail we have blog posts:

Woooo 🎉 great to see Strimzi maintainers Kate Stanley and Paolo Patierno on stage at #Current25 talking about how Strimzi helps you migrate your K8s Kafka clusters from Zookeeper to KRaft

After a very interesting lunch time at #Current25 speaking to folks from OpenAI about Mirror Maker and what we could do in Strimzi to make things easier, I'm watching the Flink AMA MC'd by @rmetzger.bsky.social

My ⚡ talk choice at #Current25 is @hpgrahsl.bsky.social and vector stores in Flink

Ok so the main thrust is actually about removing Kafka from the mix & what you need to do in Iceberg to emulate topic compaction. A really interesting custom solution due to missing features in the Flink Iceberg connector. The headline, it works, it doesn't need Kafka but it's not as fast #Current25

Onto something a bit more concrete now at #Current25. Improving CDC performance using Kafka, Flink and Iceberg

Kicking off day 2 of #Current25 by boarding the AI agent train🤖🚂 with MongoDB 🌿

Wooop Open Source Datastack FTW! Pinterest moving from an in-house to OSS based architecture! #Current25

Final talk of the day at #Current25, Async processing with Flink SQL

Great to hear that OpenAI have identified and fixed several issues with PyFlink and intend to push those fixes upstream 👏👏👏

It's PyFlink for all the things and OpenAI uses the upstream Flink Kubernetes Operator

Learning about how OpenAI built its stream processing system using Kafka and Flink at #Current25

This was a very interesting talk full of intriguing ideas. However, there was a lot of "you _could_ do <complicated thing> and then do <even more complicated thing> based on that". I would love to see some real life demonstrations of all these agents linked together doing what we expect.

This talk has some really great examples of Flink's complex event processing (CEP) functionality in Flink SQL

Great to see a push to enhance upstream Flink SQL with addional AI orientated functions

Next up at #Current25 Flink 🐿️ but with robots 🤖 (AI agents)!

Next at #Current25 a panel discussion on the past, present and future of @kafkastreams.bsky.social. I love the background...the likeness is uncanny...

Starting #Current25 with an overview of anomaly detection in Flink SQL. This is from Confluent and AFAICS uses their proprietary ML functionality. But there is work in the upstream to support the same that builds on FLIP-437 which added the model syntax to Flink SQL

cwiki.apache.org/confluence/d...

Ready to start another Kafka and Flink filled couple of days in London at #Current25

Looking forward to hearing about stream processing with Flink, integrating with open table formats and this newfangled AI thing...