Health and observability
Readiness, correlation, trace-safe logging, scheduler evidence, and projection freshness.
Health surfaces
Actuator health and information endpoints are public by explicit security configuration. The management surface itself answers to its configured permission for sensitive operations.
Readiness includes whether the deployment has a managed station. A process may be alive while the station application is not ready to serve operational work.
Request correlation
Every request receives or propagates correlation context. Integration events carry correlation and causation identifiers so an operator can follow one business action across modules and asynchronous consumers.
Trace-safe logging
@Traced records arguments, result, duration, and failure without logging
arbitrary DTO contents. Values are silent unless their type explicitly opts
into a safe summary.
Warnings remain on business rejection paths where the code can state why a decision was refused. Sensitive bank, identity, credential, and personal data must never be made log-safe for convenience.
Scheduled jobs
Every scheduled method carries a distributed lock with explicit maximum and minimum duration. PostgreSQL grants and times the lock, avoiding pod clock skew.
lockAtMostFor must exceed the slowest possible job run. Durable domain dedup
protects one-time announcements even when a run exceeds its lock.
Projection freshness
Asynchronous projections should expose freshness or last-applied-event evidence when a user relies on them. Kafka delivery is at least once; consumers must be idempotent and should make lag visible rather than presenting stale data as current truth.