Queues, Kafka & Resilience Patterns
Question bank: queues vs logs, Kafka partitions and the KIP-848 rebalance protocol, lag that lies, DLQ redrive, the real scope of exactly-once, retry budgets, deadlines, breakers, backpressure, shedding and degradation ladders.
About 110 minutes · 26 exam questions · 25 interview questions · 2 code labs
What you will learn
- Queues — work distribution vs replayable log
- Kafka — partitions are the unit of order, parallelism and pain
- Kafka rebalancing — classic protocol vs KIP-848
- Consumer lag — the metric that can lie
- Dead Letter Queue — a parking lot with an owner
- Exactly-once semantics — what Kafka actually guarantees
- Retries — transient only, one layer, on a budget
- Timeouts — stop waiting AND stop working
- Circuit Breaker — fail fast when a dependency is clearly broken
- Backpressure — make the producer slow down
- Load shedding — refuse work you cannot finish in time
- Graceful degradation — a smaller product, on purpose
Previous: Consistency, Consensus, Locks, Sagas & Event-driven Patterns · Next: Microservices, Observability, SLOs, Releases & Disaster Recovery