Interview question. Kafka cluster A is asynchronously mirrored to cluster B. When A becomes unavailable, the application starts consumers in B but reprocesses hours of events. Explain why the same offset does not mean the same record in two clusters, then design a safe cutover, consumer-position recovery and eventual failback.
Queues, Streams & Event Processing · Principal
How would you recover Kafka consumers after failing over to a mirrored cluster in another region?
Designing Reliable Distributed Messaging Systems · Part 6 of 6 · 5 min read
Read the worked explanation ↓Part of paid interview preparation
The worked explanation, deeper discussion, timed interview exercise with assessment criteria, reasoning check, transfer exercise and spoken revision answer for this lesson are available to readers with paid access.
Checking your access…
About Interview Arena →