Health dashboard

The Health tab shows system-level health proxies as a card grid, with detailed sections for specific risk areas below.

Top-level cards

Card	What it means
Cache hit ratio	Whether reads are mostly served from PostgreSQL/shared OS cache or frequently hitting disk.
Connections	Current connection usage relative to the configured ceiling.
Oldest transaction	Whether long-lived transactions may be blocking cleanup or vacuum progress.
TXID wraparound risk	Whether transaction ID or multixact age is approaching risky territory.

Exact color thresholds and scoring gates are intentionally not published. The UI shows the current value, severity, and explanation for each finding.

Checkpoint stats — requested vs timed checkpoints, WAL rate, completion target. On managed providers (Aiven, RDS, Azure, GCP), an italic note warns that checkpoint frequency may reflect provider housekeeping rather than application write pressure.
Sequences at risk — sequences approaching their max value, with CRITICAL/WARNING severity.
Replication slots — all non-temporary slots (active and inactive) with WAL lag. Inactive slots shown in alert table; active-but-lagging slots used by the Data Safety card.
pg_stat_statements capacity — eviction rate and deallocation count.
bgwriter maxwritten — fires when the background writer hits its page limit.
Vacuum horizon blockers — sessions holding old snapshots that prevent dead tuple cleanup.

Two cards assess disaster recovery posture:

Crash Recovery (RTO) — estimates how much WAL may need to be replayed after a crash.
Data Safety (RPO) — considers replication sync state, write lag, and replication slot lag.

Each finding can be snoozed locally so known, accepted conditions do not keep distracting you during the current investigation window.

Health data is short-lived cached. Stale content stays visible while fresh data loads in the background.