prisma-connection-pool-exhaustion

Fix P2024 errors by routing through a connection pooler such as Supabase's PgBouncer on port 6543 or Prisma Accelerate, and add connection_limit=1 to your DATABASE_URL. This caps each serverless instance at one connection so concurrent functions cannot exhaust the database limit.

Serverless platforms create a new Prisma client per cold start, and each instance opens multiple connections by default. Under traffic spikes, hundreds of concurrent instances exceed the managed database connection limit, while local development uses a single long-lived process.

Yes, Supabase provides a pooled connection string on port 6543 with pgbouncer=true in the DATABASE_URL. Use this instead of the direct port 5432 connection when deploying Prisma to serverless environments.

The connection_limit=1 parameter restricts each Prisma client instance to a single database connection. Combined with pool_timeout=20 and connect_timeout=10, it prevents pool exhaustion in serverless, though you can raise it if latency appears.

Use Prisma Accelerate when you want built-in connection pooling and caching without managing PgBouncer or provider-specific pooler URLs yourself. It is enabled via npx prisma generate --accelerate and suits teams preferring a managed solution.