Neon Serverless Postgres Cold-Start Timeout / Connection Exhaustion

Waking a scale-to-zero Neon compute takes ~300-800ms; short client connect_timeout (<5s) or unpooled direct TCP connections fail during wake-up.

1. Immediate Code & SKU Fix

Replace With SKU: @neondatabase/serverless HTTP driver + -pooler endpoint

Install Command: npm install @neondatabase/serverless

Required Env Vars: DATABASE_URL

import { neon } from '@neondatabase/serverless';
// Use the -pooler hostname with ?sslmode=require&connect_timeout=15
const sql = neon(process.env.DATABASE_URL!);
const rows = await sql`SELECT 1`;

2. Verified Hot-Swap Fallback Route

JTBD: serverless-db

Primary: neon → Fallback: supabase-pgvector

liveHealth: operational · autoActOk: true

Want Auth+Storage+Realtime suite — Supabase.

GET /api/rescue JSON (<2 KB)Falsify Pick (/api/falsify)

Matched Error String Signatures (5)

neon cold startneon cold start timeoutcompute time exhaustedtoo many clients alreadycanceling statement due to statement timeout

curl -s "https://getagentdepot.com/api/rescue?q=neon%20cold%20start"

Browse All 28 Agent Runtime Rescue Cards