role-database:connection-management
Connection pooling and management across all engines. PgBouncer (transaction/session/statement mode), pgcat (Rust-based, load balancing, sharding), ProxySQL (MySQL query routing, caching), HikariCP (Java), application-level pooling (Prisma, SQLAlchemy, GORM, database/sql). Serverless connection s...
SKILL.md
Full skill instructions
Connection Management
Why Connection Pooling
- PostgreSQL: ~10 MB per connection (per-process model)
- MySQL: ~1-4 MB per connection (per-thread model)
- Without pooling: N app instances × M connections = N×M database connections
- Pool reuses ~20 persistent connections to serve 100 simultaneous requests
Reference Files
Load from references/ based on what's needed:
references/pgbouncer-proxysql.md
PgBouncer pool modes (transaction/session/statement), full pgbouncer.ini config, admin console monitoring commands. pgcat Rust-based alternative with read/write routing config. AWS RDS Proxy overview. ProxySQL MySQL read/write splitting with hostgroups and query routing rules. Serverless strategy comparison table (Prisma Accelerate, RDS Proxy, Neon, Supabase, Hyperdrive). Load when: configuring PgBouncer, pgcat, ProxySQL, or serverless connection proxies.
references/app-pooling-sizing.md
Application pool configuration: node-postgres, Prisma, SQLAlchemy, HikariCP (Java), database/sql (Go). Pool sizing formula: (core_count * 2) + spindle_count with worked examples. Connection leak detection queries for PostgreSQL and MySQL. Leak prevention checklist. Load when: configuring pools inside application code or diagnosing connection leaks.