PoolAudit
A one-shot diagnostic that reads your current PgBouncer config, replays a synthetic workload, and tells you exactly how much throughput you're leaving on the table.
Backend and platform engineers about to scale a Postgres-backed service
- CLI and web app: paste your config, get a benchmark report in 10 minutes
- Compares single-process vs. fleet-of-N with the same so_reuseport + peering setup ClickHouse describes
- Generates a ready-to-apply Terraform/Helm snippet for the recommended fleet
- Free for single-process audits; paid tier for full fleet recommendations and PR-bot integration
Most teams run one PgBouncer because that's the default; the HN thread proves engineers are hungry for a quick way to know whether they're bottlenecked before they hit production stalls.
Multiple recent guides and talks (ClickHouse, pgstef PGDayLowlands, dev.to, oneuptime) confirm PgBouncer tuning/bottlenecks are an active pain point, but the cited HN thread with 93 points could not be verified via search and the audience is a niche subset of backend/platform engineers.How we scale PgBouncer in ClickHouse Managed Postgres ↗PgBouncer at scale - pgstef.github.io ↗
No dedicated one-shot PgBouncer diagnostic SaaS found; pganalyze is ongoing monitoring with tuning advisors and PGTune is a free static PostgreSQL config calculator — leaving room for a focused workload-replay diagnostic.Postgres performance at any scale | PostgreSQL Tuning - pganalyze ↗PGTune - calculate configuration for PostgreSQL ↗
pganalyze proves WTP for Postgres perf tooling, but a one-shot diagnostic is hard to monetize versus subscription monitoring; abundant free blog guides lower perceived value of a paid single-run report.Postgres performance at any scale | PostgreSQL Tuning - pganalyze ↗Boosting PostgreSQL Performance with PgBouncer: A Configuration Guide ↗
PgBouncer is mature and long-lived so the problem persists, but a one-shot audit is a utility with bounded repeat use rather than an ongoing platform.
Reading configs is easy, but safely replaying a synthetic workload against a user's pgbouncer (pool modes, auth, isolation) is non-trivial engineering beyond a simple parser.