What causes dbt Adapter Connection Pool Exhausted?

The dbt job's `threads` setting exceeds the number of warehouse connections available to the service account. The warehouse has a per-user or per-role concurrent connection limit lower than the dbt thread count. Connection leak: previous dbt runs left connections open and the pool is partially occupied by idle connections. dbt is running multiple jobs concurrently (e.g., dbt test and dbt run simultaneously) and the combined thread count exceeds the pool size

How do I fix dbt Adapter Connection Pool Exhausted?

Reduce the `threads` setting in the dbt profile or dbt Cloud environment: start at 4 and increase cautiously based on warehouse capacity.. Check the warehouse's session limit for the service account role: `SHOW PARAMETERS LIKE 'SESSIONS_PER_USER' IN ACCOUNT` (Snowflake) or equivalent.. Add a `connection_timeout` to the profile to force connection requests to fail fast rather than hang indefinitely.. For concurrent dbt jobs: stagger their schedules or reduce thread counts so the total connections across all concurrent jobs stays within the warehouse limit.. Upgrade the connection pool handling in the adapter if using an older version — recent adapters have improved pool cleanup on run completion.

High severitydbt

Power BI Refresh Error:
dbt Adapter Connection Pool Exhausted

What does this error mean?

The dbt adapter's connection pool ran out of available connections to the warehouse. New model threads cannot acquire a connection and fail with a pool exhaustion or timeout error.

Common causes

1The dbt job's `threads` setting exceeds the number of warehouse connections available to the service account
2The warehouse has a per-user or per-role concurrent connection limit lower than the dbt thread count
3Connection leak: previous dbt runs left connections open and the pool is partially occupied by idle connections
4dbt is running multiple jobs concurrently (e.g., dbt test and dbt run simultaneously) and the combined thread count exceeds the pool size

How to fix it

1Reduce the `threads` setting in the dbt profile or dbt Cloud environment: start at 4 and increase cautiously based on warehouse capacity.
2Check the warehouse's session limit for the service account role: `SHOW PARAMETERS LIKE 'SESSIONS_PER_USER' IN ACCOUNT` (Snowflake) or equivalent.
3Add a `connection_timeout` to the profile to force connection requests to fail fast rather than hang indefinitely.
4For concurrent dbt jobs: stagger their schedules or reduce thread counts so the total connections across all concurrent jobs stays within the warehouse limit.
5Upgrade the connection pool handling in the adapter if using an older version — recent adapters have improved pool cleanup on run completion.

Frequently asked questions

What is the recommended thread count for dbt Cloud production jobs?

dbt Labs recommends 4–8 threads for most production jobs. Higher thread counts (16+) are only beneficial for very large projects on large warehouses with high connection limits. Always test in non-production before increasing.