The Flink Mistake Playbook 2 Years of Real World Debugging by Naci Simsek

00:00 Intro 00:10 About Naci Simsek 01:26 The Flink Mistake Playbook 02:51 Migrating from ld to new Kafka connector 17:34 Distributing the load evenly among task slots 25:02 Avoiding performance issues of Kryo Fallback Let's be honest, working with Flink can be tricky. Even experienced folks stumble. This talk is all about learning from common pain points, so you don't have to repeat them! I will zero in on three crucial mistakes that trip up many Flink users, mistakes that can seriously impact your pipeline's performance and stability. I will be talking about: Kafka Connector Chaos: Ignoring the migration rules when upgrading Kafka connectors? Say hello to state that just keeps growing… I will show you how to migrate like a pro. Serialization Sabotage: Are you unknowingly tanking your throughput with Kryo fallback? Serialization matters a lot in Flink, and I will show you how to get it right. Load Balancing Blues: Is your load all over the place, with some TaskManagers chilling while others are sweating? I will dive into strategies for even load distribution to keep your Flink cluster happy and efficient. This is a practical, no-nonsense session packed with tips and tricks to avoid these common pitfalls and build robust, performant Flink applications.