Currently, the fecfile-api-migrator is spinning up a full instance of the fecfile-web-api using bin/run-api.sh. As configured, this results in 9 gunicorn workers with 8 threads each = 72 database connections. This is unnecessary for simply running migrations. Ticket 2990 investigates whether we need a running copy of the API at all for management commands like migrate. If we do, we should see if we can reduce the required database connections as much as possible (maybe 1 worker with 1 thread) to avoid database connection utilization spikes during deployments. If we don’t need it, we should adjust how we’re running migrate as part of deployments accordingly.
QA Notes
null
DEV Notes
null
Design
null
See full ticket and images here: FECFILE-2992