How we made PG Fitter, Happier, More Productive
summary
Postgres setup scales by adding RDS Proxy for connection pooling, tuning autovacuum (raise cost_limit, lower cost_delay) and increasing provisioned I/O and instance size, rebuilding indexes with PG Repack (30 rebuilt, ~230 GB freed) and dropping >100 unused indexes (≈300 GB freed), relocating six heavy-write metrics tables to a new database with zero downtime (≈700 GB moved), using pg_stat_statements and PG Hero to find top queries, adding statement_timeout = 5s and other timeouts to fail fast and keep the system responsive.
AI-written summary · about
chapters (5)
related talks
58mScaling Postgres
51mScaling Postgres to the next level at OpenAI
57mPostgreSQL When It's Not Your Job
37mPostgres at Any Scale
25mScaling PostgreSQL Beyond Query Optimization
36mPostgres at any Scale
40mPushing past PostgreSQL’s vertical scaling limits - Running Amazon RDS
30mPuny to Powerful PostgreSQL Rails Apps