Tag
How a duplicate-payouts report from finance taught me that idempotency is a system property, not a one-service bug fix — and the path from a tactical patch to a cross-cutting framework that took at-least-once seriously.
A "simple" lost-update bug took me through optimistic locking, pessimistic locking, and finally a two-stage Redis-locked aggregator — a tour of why the right concurrency primitive depends entirely on the shape of your contention.
Google's SRE book is 500 pages long and targets 100-engineer orgs. For a 10-person team, the pragmatic subset that delivers most of the benefit at a fraction of the cost.
What DLQs are, why you must have one for every message consumer, and the operational patterns that keep bad messages from blocking the good ones.
Why every non-trivial backend needs idempotency keys, the common designs, and the mistakes that let duplicates slip through anyway.
Why "save to DB, then publish to Kafka" is almost always wrong, and the outbox pattern that fixes it — with real Java code, schema, and production considerations.