Postgres FM pg_ash
Feb 20, 2026
Nikolay Samokhvalov, founder of Postgres.ai and creator of pg_ash, builds lightweight PostgreSQL tooling for observability and performance. He introduces pg_ash, its low-footprint, non-extension design, and compact per-second sampling approach. Discussion covers storage rotations, dictionary management, pg_cron scheduling, and plans for retention and rollups.
AI Snips
Chapters
Transcript
Episode notes
Wait-Event Analysis As The Primary Lens
- Wait-event (active session) analysis is the most revealing single approach for database troubleshooting.
- Nikolay says it should be the starting point for understanding Postgres performance issues.
Prefer Internal Sampling Over External Polling
- Use internal sampling to avoid external pull frequency limits and reduce observer effect.
- Aim for a low-footprint, in-Postgres sampler rather than relying solely on external monitoring.
Built As An "Anti-Extension"
- Nikolay built pg_ash as an "anti-extension": pure SQL/PLpgSQL so it runs on managed platforms.
- He chose this after finding extensions are often unavailable on managed services like many hosted Postgres offerings.