Partitions, Keys, and Ordering
Design a small subsystem that relies on Partitions, Keys, and Ordering. Outline components, data flow, failure modes, and metrics.
Answers use simple, clear English.
Quick interview answer
Use Partitions, Keys, and Ordering as the core idea. Example shape: Order events keyed by orderId so all status updates for one order stay ordered..
Detailed answer
Use Partitions, Keys, and Ordering as the core idea. Example shape: Order events keyed by orderId so all status updates for one order stay ordered.. Watch for: Cross-partition ordering impossible; hot keys imbalance partitions.. Measure success via latency/error/saturation. Core: Ordering guaranteed per partition only. Key determines partition hash — same key → same partition. More partitions = more parallelism but more overhead. Real-time example: Order events keyed by orderId so all status updates for one order stay ordered. Pros: Horizontal consumer scale; replay by offset. Cons: Cross-partition ordering impossible; hot keys imbalance partitions. Common mistakes: Assuming global order; too few partitions limiting throughput. Best practices: Choose key for locality; monitor per-partition lag; plan partition count upfront. Audience level: Junior.
Full explanation
Ordering guaranteed per partition only. Key determines partition hash — same key → same partition. More partitions = more parallelism but more overhead.
Real example & use case
Order events keyed by orderId so all status updates for one order stay ordered.
Pros & cons
Pros: Horizontal consumer scale; replay by offset. Cons: Cross-partition ordering impossible; hot keys imbalance partitions.
Common mistakes
Assuming global order; too few partitions limiting throughput.
Best practices
Choose key for locality; monitor per-partition lag; plan partition count upfront.
Follow-up questions
- How would you test Partitions, Keys, and Ordering?
- What metrics prove Partitions, Keys, and Ordering is healthy in prod?
- How does Partitions, Keys, and Ordering change at 10× traffic?