CircleCI outage: Postgres disk filled by a scheduled cleanup job

On October 9, 2026, CircleCI halted pipeline processing for about 12 hours after a scheduled Temporal history-deletion job filled its 1TB Postgres log volume. The fix shipped after a similar October 6 outage did not account for the second, already-scheduled load.

• The October 9 outage stopped CircleCI pipelines for roughly 12 hours
• Log volume free space fell from 0.8TB to 0.1TB in about two hours
• Cleanup tables later held tens of millions of rows, slowing all operations
• Only pausing new and retried work let the database catch up