Investigating consumer lag without guesswork
Hi everyone, we’ve created a new project called Broka that supports Kafka, RabbitMQ, Redis, ActiveMQ Artemis, and Memcached. If you don't mind, I’d love to walk you through it with a quick guide. You can also check it out at broka.dev — would love to hear your thoughts and feedback! Let me start introduction <3 A Kafka consumer group shows a lag of 0. Is it caught up? Not necessarily. Offset lag…
A new project called Broka has been developed, which supports Kafka, RabbitMQ, Redis, ActiveMQ Artemis, and Memcached. The creators are inviting individuals to explore the project at broka.dev and provide their feedback. During their presentation, they discussed consumer lag in Kafka and explained that a lag of 0 does not necessarily mean a consumer is caught up.
They clarified that offset lag only counts the partitions a group has committed to, and a brand-new group that has never committed anything will also display a lag of 0. The difference lies in the state beside the number: if members hold assignments, it is considered stable, indicating a backlog that someone is working on. If no consumers are present, the lag will not move until a consumer joins, regardless of the size of the backlog.
Lastly, if the number is changing due to an assignment that is being rebalanced, it signifies that the group is in the process of redistributing tasks.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
