conduktor.io ↗

KIP-441 — Smooth Scaling Out for Kafka Streams

Accepted Kafka 2.5 StreamsConsumer

Introduces a `warmup` task state in Kafka Streams where a new instance restores standby changelogs before being assigned active tasks, so the prior active owner retains the active assignment until the new owner is sufficiently caught up. After a rebalance, stateful tasks assigned to a new instance could block processing for hours while rebuilding their state stores from scratch, causing severe availability degradation during scale-out.

Details

AuthorGuozhang Wang
StatusAccepted
Kafka Version2.5
JIRAKAFKA-8019
WikiView on Apache Wiki
Created2019-03-15
Last Modified2020-05-01
Explore how this KIP affects the Kafka protocol in the Protocol Explorer, or see the full KIP database.