Managing a Kafka cluster in KRaft mode with Strimzi
summary
Strimzi shows how to run Kafka in KRaft mode, automating a five‑phase migration from ZooKeeper while requiring the user to create a KafkaNodePool for controllers and trigger/finalize the process via a CR annotation. Recommended controller replica count matches the original ZooKeeper count, and rolling order is explicit: restart unready pods, follower controllers (and combined followers), then the active controller, then brokers. Key metrics to monitor include active_controller_count == 1, election_latency and last_applied_record_lag_ms during the dual‑write phase.
AI-written summary · about
chapters (9)
- 0:00Kafka KRaft mode intro
- 1:04What is Kafka KRaft mode
- 5:20Controller and broker roles
- 7:47KRaft cluster metrics
- 12:01Controller metrics in KRaft
- 15:57Rolling updates in KRaft
- 18:44Migrating Zookeeper to KRaft
- 27:16KRaft with Strimzi summary
- 28:19Q&A
related talks
26mReleased From the Cage: Apache Kafka Without Its ZooKeeper
31mSimplifying Apache Kafka on Kubernetes With Strimzi
24mKafka & Kubernetes - Migrating to a World Without Zookeeper
24mMake Your Kafka Cluster Production-Ready
29mMastering Kafka Workload Balancing With Strimzi’s Cruise Control Integration
26mProduction Grade Kafka on K8s
30mFrom Running to Operating: Real-World Strimzi in Production
42mTransition to Apache Kafka on Kubernetes With Strimzi