Database connection pooling — configuration and sizing

Application creates too many database connections (resource exhaustion) or too few (request queuing). Connection pool misconfiguration causes intermittent timeouts, slow queries, or database overload.

Pool sizing formula: connections = (core_count 2) + effective_spindle_count. For SSD: ~(cores 2) + 1. (1) Start small: 10-20 connections per application instance. (2) Set idle timeout: close connections that sit idle > 10 minutes. (3) Set connection lifetime: recycle connections after 30 minutes to handle DNS changes and server-side limits. (4) Set acquisition timeout: how long to wait for a connection from pool (5-10s, fail fast). (5) Monitor pool metrics: active connections, idle connections, wait time, timeouts. (6) For serverless: use external pooler (PgBouncer, RDS Proxy) because each invocation may create a new connection. (7) Test under load: pool exhaustion only surfaces during traffic spikes.

Each database connection consumes memory on both the app and database server (~10MB for PostgreSQL). Too many connections cause context switching overhead on the database. Too few cause request queuing.