Backend Reliability

Connection Pool Sizing and Timeout Rules Prevent Database Exhaustion

April 3, 2026 7 min read

An undersized pool starves traffic. An oversized pool overwhelms the database. Pair pool config with query timeout limits and health checks on acquire.

Rule

Database connection pools must be sized based on measured concurrency with bounded timeout on acquire.

Why

Unbounded pools or no acquire timeout causes database overload and cascading failure under traffic spikes.

Must

Set a maximum pool size informed by the database max_connections limit.
Set an acquire timeout to fast-fail rather than queue indefinitely.
Set a query timeout at the connection level as a last-resort safety net.
Test with connection at max and verify graceful degradation.

Should

Set minimum idle connections to remove cold-start latency.
Monitor pool utilization and alert at 80% sustained usage.

Anti-patterns

Pool max set to unlimited.
No acquire timeout, causing requests to hang indefinitely.
Creating a new connection per request outside the pool.

Test Cases

Saturated pool returns 503 within the acquire timeout window.
Pool recovers automatically after a connection spike.

Telemetry

pool_utilization_percent
pool_acquire_timeout_errors
pool_acquire_latency_p99