Engineering Notes

Deep-dives into decisions, incidents, and lessons from building data platforms and backend infrastructure at scale.

Published

The 50M-row Debezium snapshot incident

A snapshot operation consumed 60% CPU and nearly blocked a critical migration. How we recovered and what it taught about production risk vs architectural purity.

Incident record

Ten production problems covering diagnosis, architecture, customer judgment, correctness, resilience, and leadership under pressure.

Read all ten incident records →