At a glance
Active Sessions is the live count of open Snowflake sessions: connections held by users, BI tools, drivers, and scheduled jobs that are currently authenticated to the account. It is the connection-side view of demand. A flat baseline that matches your known client fleet is healthy; a sudden climb usually means a connection leak in an application pool or a BI tool that has stopped closing sessions cleanly.
| What it tracks | The number of currently open sessions on the account for the selected period, read from SESSIONS in the ACCOUNT_USAGE / INFORMATION_SCHEMA views and the live session list. |
| Data source | detail: Active Sessions for the selected period. Counts authenticated sessions, not running queries. |
| Time window | RT (real-time, refreshed on the live polling cycle). |
| Alert trigger | None configured by default. Pair with saturation and queue-depth cards for demand alerting. |
| Roles | owner, platform, SRE |
Calculation
Calculated automatically from your Snowflake account state. See the At a glance summary above for what the metric tracks and the reconcile pointer below for the native check.What it tracks
A session in Snowflake is an authenticated connection to the account. Every BI dashboard, dbt run, JDBC/ODBC client, Python connector, and Snowsight worksheet opens at least one session. Sessions are distinct from queries: a session can sit idle for a long time without running anything, and a single session can run thousands of queries over its life. Counting active sessions tells you how many clients are connected, which is the leading indicator of connection-pool health rather than compute spend. A creeping session count with flat query volume is the classic signature of a leaking connection pool in an upstream application, where sessions are opened but never released. Read this card with Queries per Hour (live) to separate “more clients” from “more work”, and with Failed Logins (24h) when an unexpected session surge looks like it could be unwanted access. Reconcile against the source: querySELECT COUNT(*) FROM SNOWFLAKE.ACCOUNT_USAGE.SESSIONS WHERE DELETED_ON IS NULL; for the account view (note ACCOUNT_USAGE latency), or use SHOW SESSIONS for the live list. The Vortex IQ count should match the live open-session count at poll time.