Deep Database Collection Did Not Need Copied Passwords

Copying every application database password into the observability stack would have expanded the secret blast radius just to collect metrics. On the finished hserver stack, collector executes reviewed queries inside existing database containers is the signal that makes the difference visible. The hserver collector reuses the database container's local runtime context instead of creating a second inventory of application credentials in monitoring.

The engineering pattern here is least-privilege observability design. Good monitoring should shorten diagnosis, so I prefer a small number of signals with clear semantics over a larger collection whose meaning is unclear during a failure.

Operationally I keep this constraint: Prefer narrow local collection paths, document the required Docker access, and avoid duplicating secrets unless an independent exporter truly needs them. It gives the dashboard, alert, and runbook the same interpretation instead of letting each layer invent its own definition of healthy.

The implementation can be traced to hserver commit b65d5d4. That provenance is part of the article because these notes document an actual production observability system, not a hypothetical monitoring design.