How we keep Azure Database for PostgreSQL free of bloat to maximize disk space
summary
Automation runs weekly to read table stats into a central metadata DB, call Databricks analysis, segment tables into seven clusters by reltuples and dead_tuples, and update per-table autovacuum knobs (notably autovacuum_vacuum_cost_limit increased and autovacuum_vacuum_cost_delay reduced) via an Airflow DAG. The 7-step pipeline reclaims about 1.5 TB reclaimed of server disk across clients and yields ~70% faster average query times, proven on Azure Database for PostgreSQL Flexible Server and fully automated and monitored.
AI-written summary · about
chapters (12)
- 0:00Introducing Bob and Eleni
- 1:00Talk begins: Keeping Azure Database for PostgreSQL free of bloat
- 2:31About Ebiquity
- 3:12Challenges: scope change, deletes, inserts, and bloat
- 4:57Client workload types
- 6:21Ineffective autovacuum settings cause massive bloat
- 8:13Structural performance improvements
- 9:25Design principles: automation, monitoring, and vacuum
- 12:06End-to-end pipeline in Airflow (DAG)
- 16:10Database table segmentation
- 18:52Bloat & vacuum learnings
- 21:39Q&A
related talks
34mData modeling for those who want to save time and money!
27mBoosting Azure Database for PostgreSQL Performance with Azure Advisor
25mWhat’s New with Azure Database for PostgreSQL- Flexible Server
28mDear Azure Database for PostgreSQL, can you automate my index?
26mWhat Enterprises like about Azure Database for PostgreSQL - Flexible Server
56mSeamless Migration to Azure Database for PostgreSQL
20mBuild Your Business Based On PostgreSQL Databases
32mWhat makes Azure Database for PostgreSQL Great for Developers?