Documentation ›Operations
OPERATIONS / TECHNICAL

Actionable improvement checks

How Observe distinguishes verified problems from missing monitoring and incomplete history.

Repaired sources

The PostgreSQL check uses pg_up scoped to the backend PostgreSQL exporter; the unexported postgresql@18-main systemd label previously made every dependent application unverified. Backend and VPS backup freshness use existing exporter contracts. PostgreSQL restore-drill and RabbitMQ recovery-stage gauges supply scoped recovery evidence. Backend, VPS and Mookie restart checks use hourly counter increases. Receiver coverage uses its collector timestamp.

Probes and logs

A registered nutsnews-backend readiness probe now supplies response time and freshness. HTTPS probes use independent TLS connections so session reuse cannot drop certificate-expiry evidence. NutsNews Loki inspections select host rather than the absent instance label; Raspberry retains its project scope.

Policies and presentation

Weekly local image backups warn after eight days and fail after nine. Daily local database cloud backups warn after 30 hours and fail after 48. Scheduled ingestion uses the existing 15-minute exporter contract, scoped to production backend ownership. Unknown remains the conservative machine status; evidenceState provides not-configured, no-data, collecting-history or policy-required presentation. Gaps earn zero health credit and no coverage. Probe-only coverage requires every registered probe and cannot override missing configured metrics.

ON THIS PAGE

THREE WAYS TO UNDERSTAND

Same system.
Your level of detail.

Switch reading modes above to find the explanation that fits.

Explore Expert →

Keep exploring

Public observability workspace

Public native dashboards, server navigation and owner changes.

↗
Backups & recovery

Protect existing data and know which recovery path to use.

↗