At a glance
Connections In Use is the count of open SQL client connections across the cluster right now. It is the raw numerator behind connection-pool health: every application worker, BI tool, and migration job that holds a session to CockroachDB shows up here. Reading it next to your configured maximum tells you how much connection headroom is left before clients start queuing or being refused.
| What it tracks | Connections In Use for the selected period, summed across all live nodes. |
| Data source | The sql.conns time-series metric (open SQL connections per node), also visible in the DB Console SQL dashboard “Open SQL Sessions” graph and in SHOW SESSIONS / crdb_internal.cluster_sessions. |
| Time window | RT (real-time, refreshed on each poll). |
| Alert trigger | None on this card directly. The saturation ratio is alerted on Connection Pool Saturation % and Connection Pool at >90% Saturation. |
| Roles | DBA, platform, SRE |
What it tracks
This card reports the total number of established SQL connections held against the cluster at poll time, aggregated from the per-nodesql.conns gauge. Each connection consumes memory and a slot against the cluster setting server.max_connections_per_gateway (and, on CockroachDB Cloud, against the plan’s connection ceiling). The count rises with application concurrency, connection-pool sizing in the client (for example a pgbouncer or HikariCP pool), and any long-lived analytics sessions. Watched on its own, Connections In Use is a workload signal: a steady climb usually means a pool is leaking or a new service has been deployed without right-sizing its pool. Watched against the maximum, it becomes the early-warning input to saturation: connections in use divided by the configured limit is the saturation percentage that the pool alert cards trip on.
Reconciling against the source
To verify the live figure natively, runSELECT count(*) FROM crdb_internal.cluster_sessions; or SHOW SESSIONS; for the per-session detail, and compare against the DB Console SQL dashboard “Open SQL Sessions” panel. On CockroachDB Cloud, the Metrics tab plots the same sql.conns series. Small discrepancies are normal because sessions open and close continuously; sample over a few polls rather than trusting a single instantaneous reading.