When a deployment left two databases running
A failed deployment left old and replacement database and broker services running together. Recovery meant reconciling their state, replaying accepted orders, and changing how stateful services update.
Things I encounter while building software: the problems, the decisions, and what I learn along the way.
A failed deployment left old and replacement database and broker services running together. Recovery meant reconciling their state, replaying accepted orders, and changing how stateful services update.
A misunderstood estimated-age input produced future birth dates and negative ages. Fixing the resulting warehouse failures took clearer inputs, server validation, and careful repair of existing records.
Introducing a broker left a gap between committing application state and publishing a task. A transactional outbox made the intent to dispatch durable.