feat: production-tuned pool defaults + tuning guide #168

Merged

ash merged 1 commit from feat/production-tuning into main

2026-03-02 22:48:51 +00:00

ash commented

2026-03-02 22:48:33 +00:00

Owner

pgstore.NewPool with production defaults + docs/production-tuning.md covering pools, checkpoint batching, Postgres server settings, TCP keepalive for failover.

ash added 1 commit

2026-03-02 22:48:33 +00:00

feat: production-tuned pool defaults + tuning guide — Refs #166

CI / test-sqlitestore (push) Failing after 4s

Details

CI / test-pgstore (push) Failing after 4s

Details

CI / test-nats (push) Failing after 5s

Details

CI / test-integration (push) Failing after 4s

Details

CI / benchmarks (push) Failing after 4s

Details

CI / test-codec-otelkit (push) Successful in 12s

Details

CI / test-sqlitestore (pull_request) Failing after 5s

Details

CI / test-pgstore (pull_request) Failing after 5s

Details

CI / test-nats (pull_request) Failing after 7s

Details

CI / test-integration (pull_request) Failing after 4s

Details

CI / test-codec-otelkit (pull_request) Successful in 12s

Details

CI / benchmarks (pull_request) Failing after 4s

Details

CI / test-core (push) Failing after 1m13s

Details

CI / test-core (pull_request) Failing after 1m18s

Details

0604c7b96b

pgstore.NewPool: production-tuned connection pool with sensible defaults:
- MaxConns: max(4, NumCPU*2) — scales with hardware
- MinConns: 2 — no cold-start latency
- MaxConnLifetime: 30min — recycles after Postgres restarts
- MaxConnIdleTime: 5min — returns idle resources
- HealthCheckPeriod: 30s — fast dead connection detection
- WithSchema option for search_path

docs/production-tuning.md:
- Pool sizing for advisory locks
- Checkpoint batching (100x write reduction)
- Subscription tuning (batch size, poll interval)
- Full recommended production config
- Postgres server settings (WAL, autovacuum, TCP keepalive)
- TCP keepalive for faster leader election failover