At a glance
Instance Uptime is the number of seconds since the PostgreSQL postmaster last started, read from pg_postmaster_start_time(). It is the simplest liveness check on the instance: a value that keeps climbing means the server has stayed up, while a sudden drop back towards zero means the postmaster restarted, planned or otherwise. The value is sampled in real time, so an unplanned restart shows up within one refresh cycle.
What it tracks
The card computesnow() - pg_postmaster_start_time() and reports the result as an elapsed duration. As long as the postmaster process keeps running the number rises monotonically. A drop is the signal that matters: it means the instance bounced, whether from a planned maintenance restart, an out-of-memory kill of the postmaster, a crash-recovery cycle, or a managed-service failover that promoted a different node. Because the card has no alert threshold of its own (time window: RT), it is a context number you read alongside the failover-readiness and health-score cards, which carry the alerts. An uptime that resets without a corresponding maintenance window is worth investigating: pair it with Failover Readiness and PostgreSQL Health Score to confirm the cluster recovered cleanly.
Reconciling against the source
Reconcile directly withSELECT pg_postmaster_start_time(); on the instance, or SELECT now() - pg_postmaster_start_time() AS uptime; for the elapsed value. On managed services the same restart event appears in the event log: the RDS / Aurora events list (or the CloudWatch DatabaseConnections drop), the Cloud SQL operations log, or the Azure Database for PostgreSQL activity log. Note that pg_postmaster_start_time() tracks the postmaster, not the host: a managed-service failover that promotes a standby resets this value even though the underlying VM may have been running for far longer.