Database Observability: Health Model
The module builds a PostgreSQL infrastructure health model in Service Monitor Toolkit. Services are created from the assets that the module sends to Inventory. The root service DB Observability groups the clusters, and each cluster contains its associated instances.
Health Metrics
The module installs eight global metrics:
| Metric | What It Measures |
|---|---|
Instance Availability | whether recent signals have been received from the instance |
Connection Saturation | the share of occupied connections relative to thresholds |
Replication Lag | replica delay relative to thresholds |
Wait Time Share | the share of time spent waiting in database activity |
Cache Hit Shortfall | the deviation of the cache hit ratio from the norm |
Wraparound Risk | the share of XID capacity consumed |
Autovacuum Lag | the accumulation of dead rows relative to a threshold |
Database Errors | the frequency of PostgreSQL errors in the log |
Threshold-based metrics read the db_thresholds lookup and map their values to a common scale. Static ranges are used for wait time share and errors. After installation, adaptive ranges can be calculated for these metrics in the Service Monitor Toolkit interface.
Seven metrics are linked to instances. Autovacuum lag is linked to the cluster because a physical replica does not maintain table statistics suitable for this assessment.
Service Card
The service card shows the object's state, associated metrics, and the contribution of child services to the overall health score.
Analytics Center
The Analytics Center includes all services in the model and lets you move from the overall infrastructure state to an affected cluster or instance.
