Kafka and Debezium at trivago
summary
Trivago explains building a stream-first platform using Debezium CDC to publish database changes into Kafka topics organized per aggregate, so application teams stop polling databases and instead own derived views. The talk covers running this in production across regions with legacy master–slave databases, implementation details like schema handling and connector configuration, operational lessons about offsets and consistency, and real-world throughput of millions of events per day powering real-time features and analytics.
AI-written summary · about
chapters (11)
- 0:00Kafka at trivago
- 1:20Platform services with Kafka
- 4:09Clickstream analytics
- 8:51Legacy system challenges
- 13:16Race conditions and consistency
- 17:43Kafka as message service
- 21:52Kafka and Debezium value
- 24:11Debezium replication
- 26:32Kafka Streams performance
- 27:51Data ownership and quality
- 30:38Consumer app independence
related talks
42mData Streaming for Microservices using Debezium
15mManaging Data Consistency among Microservices with Debezium
41mChange Data Streaming Patterns in Distributed Systems
49mChange Data Capture Pipelines with Debezium and Kafka Streams
26mDebezium Server - New CDC runtime on the horizon!
48mStreaming Database Changes with Debezium
42mWhy PostgreSQL people should care about Kafka and Debezium?
41mApache Kafka and ksqlDB in Action: Let’s Build a Streaming Data Pipeline!