mirror of
https://github.com/neondatabase/neon.git
synced 2026-01-14 17:02:56 +00:00
ceacc29609
## Problem See https://neondb.slack.com/archives/C04DGM6SMTM/p1736526089437179 In case of queries index scan with LIMIT clause, multiple backends can concurrently send larger number of duplicated prefetch requests which are not stored in LFC and so actually do useless job. Current implementation of index prefetch starts with maximal prefetch distance (10 by default now) when there are no key bounds, so in queries with LIMIT clause like `select * from T order by pk limit 1` compute can send a lot of useless prefetch requests to page server. ## Summary of changes Always start with minimal prefetch distance even if there are not key boundaries. Related Postgres PRs: https://github.com/neondatabase/postgres/pull/552 https://github.com/neondatabase/postgres/pull/551 https://github.com/neondatabase/postgres/pull/550 https://github.com/neondatabase/postgres/pull/549 Co-authored-by: Konstantin Knizhnik <knizhnik@neon.tech>