Deployment architecture — multi-region DR

3 h
requires KFK-11

From your world. a DR runbook with RPO/RTO targets, like you'd write for a warehouse or data mart

Prompt A. "Define RPO and RTO. Then explain why an active/passive MirrorMaker 2 design has a non-zero RPO that equals the replication lag, and what determines its RTO. Contrast it with a stretched cluster. Link it to a warehouse DR runbook that you would actually write."

Prompt B. "A stakeholder wants 'zero data loss and instant failover, cheaply, across two regions'. Explain which parts of this request conflict. Explain what you would actually commit to, with numbers."

Rubric — a strong explanation covers these points (5–7):

RPO = tolerable data loss (time). RTO = tolerable downtime (time).
States that MM2 is asynchronous, so RPO = replication lag (never 0).
Says that cutover automation decides most of the RTO for active/passive.
Contrasts a stretched cluster: synchronous, RPO ≈ 0, higher latency, needs ≥3 failure domains.
Mentions offset translation, so that failed-over consumers resume correctly.
Notes that the topic prefix prevents loops in active/active.
Frames the whole topic as a trade of cost/complexity vs RPO/RTO. Notes that "0 RPO + instant failover + cheap + 2 regions" is not possible at the same time.