Server Dashboard
functionalLive health tiles for a connection - sessions, throughput, health, and resources - refreshed on a ticker.
The Server Dashboard is a live, read-only health view for a single connection. It reads catalog and dynamic-management snapshots on a five-second ticker and lays them out as metric tiles, so the state of the instance is legible at a glance rather than buried in a dozen queries you have to remember. Each tile shows the metric, its current value, and a sparkline of the last five minutes so a trend is obvious before it becomes a problem.
- Sessions
- Connections (against the configured maximum), running requests, blocked count, idle-in-transaction, longest query, and oldest open transaction.
- Throughput
- Transactions, queries, rows read, and slow queries per second - computed from the engine's cumulative counters between refreshes.
- Health
- Deadlocks and lock waits per second, temp spills, and replication lag, the numbers that tell you something is fighting.
- Resources
- Cache hit ratio, page life expectancy, memory in use, storage and reclaimable space, object counts, and uptime.
It is non-modal and there is one dashboard per connection, so you can open several and watch multiple servers side by side; each new pane cascades slightly so none hides another. Refresh and pause act only on their own pane, and a status line shows how fresh the numbers are. Footer buttons jump straight to Activity Monitor, Query Insights, and Disk Usage for the same connection, and a per-database sizes grid fills the bottom of the pane.
VIEW SERVER STATE grant), only its tiles disappear, and the pane names what was unavailable instead of pretending the number is zero.
Tiles adapt to the engine - PostgreSQL adds idle-in-transaction, shared buffers, and replication lag; MySQL and MariaDB add slow queries (read from their respective status tables); ClickHouse reads system.metrics and system.events; Oracle reads V$SYSSTAT and V$SGA; Redis reads INFO. SQLite, libSQL / Turso and DuckDB, being files rather than services, show only the file-level Resources tiles: storage, reclaimable space, and object counts.
Your own widgets
The built-in tiles cover what every server has in common; the questions only your system asks - orders waiting to ship, queue depth, last night's ETL row count, belong on the dashboard too. Any saved chart that was saved against this connection can be pinned as a widget. Click Add widget in the dashboard footer and pick from the connection's saved charts; the chart appears in a Widgets band between the health tiles and the sizes grid, drawn with the same chart renderer the results grid uses. A widget runs on the database its chart was saved against, over the dashboard's own connection route.
Each widget keeps its own refresh interval: click the interval button on the tile to cycle through manual, every 30 seconds, 1 minute, 5 minutes, and 15 minutes. The interval belongs to the saved chart, so the Chart Dashboard follows the same cadence. A widget never starts a second run while one is still in flight, Refresh re-runs it on demand, the arrow buttons reorder the band without re-running anything, and the cross removes the pin (the saved chart itself stays in your library). Pins are remembered per connection, and a chart deleted from the library quietly drops off every dashboard it was pinned to. If the pin list or the chart library cannot be read, the band says so rather than showing an empty dashboard, and the file is left untouched.
SELECT or WITH … SELECT, optionally after a USE that switches database), the same rule the Chart Dashboard applies. A saved chart whose SQL would modify data shows why it was refused instead of running.