Connection pooling

Overview

Every serverless invocation that opens its own Postgres connection competes for a fixed ceiling. lumioguard reads your database clients, runtime targets, and pool configuration to find the handlers that will exhaust it under concurrency.

What it looks for

• The pg driver used in an edge runtime route
• The direct endpoint used instead of the -pooler host in serverless
• new Pool or new Client constructed inside the handler body
• Pool max size above 10 in a serverless context
• @neondatabase/serverless missing where the edge runtime needs it

Why it matters

This fails only under load, which means it fails in production and not in staging. The node-postgres driver in an edge route, or a new Pool inside the handler body, turns a traffic spike into "too many connections" for every user at once.

Rules in this check

How the fixes arrive