446 bytes, text/plain
Apply quick fix to Sync SQL Alchemy configuration while Bug 692383 is being addressed. The installed version of SQL Alchemy, 0.6.6-1, behaves like the one here: http://stackoverflow.com/questions/5549940/sqlalchemy-connections-arent-closed-when-pool-is-overflowed The quick fix is to set pool_size += 10, and max_overflow to 0. This will keep the DB connection pool per gunicorn worker where it is currently, without allowing unused established connections to accumulate.
Created attachment 722550 [details] Diff document displaying Quick Fix change for sync.conf
Comment on attachment 722550 [details] Diff document displaying Quick Fix change for sync.conf This seems sensible to me, no need for an overflow if we're constantly operating at capacity anyway.
Comment on attachment 722550 [details] Diff document displaying Quick Fix change for sync.conf Looks fine here.
This change attempted in stage with no result: http://pencil.scl2.stage.svc.mozilla.com/dash/stage/syncstorage_details/socket_counts?from=1362804540&until=1362805491 It appears that setting max_overflow to 0 does not resolve the problem. Closing this bug.