Lesson 16 · Senior Kafka practice
Reviewing Kafka designs like a senior engineer
How to judge whether a Kafka design is truthful about durability, liveness, ordering, and operational trade-offs.
Your win: review Kafka designs for hidden trade-offs, weak assumptions, and false confidence instead of stopping at “it uses Kafka.”
Why Kafka review is its own skill
A design can use all the right words — partitions, retries, idempotency, DLQ, exactly-once — and still be misleading. Senior review is the skill of checking whether those words connect to real guarantees.
That is why this page matters. Many weak Kafka designs do not fail because the author forgot terminology. They fail because the claimed guarantee is stronger than the implementation actually supports.
What strong review comments usually target
- ordering claims that do not match the key or partition strategy
- durability claims that ignore
acksreality - retry flows that never explain poison-message handling
- idempotency assumptions with no concrete dedupe boundary
- consumer-scaling claims that ignore partition count or hotspot risk
The best review comments make the hidden contract explicit. They turn “Kafka will handle it” into “which exact behavior are we relying on, and where is that behavior enforced?”
RequireOne, AlwaysCommit(), lane design, retry loops, and missing DLQ behavior.Use the repo map as a review anchor
The strongest Kafka reviews in this repo start by checking claims against the actual wrapper behavior and service usage captured in the map.
→ Repo Kafka map
→ docs/kafka/lessons/0011-failure-modes.html
Check yourself (from memory)
Q1. What is the most senior review question for a Kafka design?
Sources. Repo Kafka map; existing Kafka lessons; curated design resources.