New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
Consumers keep dropping out and rejoining, throughput craters, and nobody touched the code.
What you'll have at the end
A consumer group that holds Stable through a full hour of checks, with flat lag and a zero rebalance count, reached by resizing the poll interval and batch to match measured processing time and replacing the assignor that was turning each eviction into a group-wide stall.
You need
A Kafka consumer group already running against a real topic in a shared or production environment, with access to its application logs and the ability to check group state and metrics against the cluster.
Not covered
One member repeatedly crash-looping on a single bad message while the rest of the group stays calm: that's a poison pill, not a poll-timeout or assignor problem, and needs a dead-letter topic instead.
One bad Kafka message stalls a partition with no dead-letter topic
go there instead if only one member keeps leaving and rejoining while the rest of the group stays calm: that's a poison pill crash-looping one consumer, not a poll-timeout or assignor problem.
Commit Kafka consumer offsets so a crash doesn't skip or repeat messages
worth doing once this settles, so a rebalance handoff you can now trust also commits cleanly instead of replaying or skipping work.
Checked 25 Aug 2026
Part of the Apache Kafka cookbook