7 or so deadly MySQL performance sins
summary
Seven common MySQL mistakes with concrete fixes: over‑allocating per‑session buffers causes memory blowouts and the OOM killer, calculate RAM as (per_session_buffers × max_connections) + buffer_pool, set vm.swappiness=0 instead of removing swap. Query cache usually hurts; set query_cache_type=0 and query_cache_size=0. Use appropriately sized primary keys (switch long 32‑char UTF‑8 IDs to integers or unsigned where possible) to shrink all secondary indexes. Verify replication with pt-table-checksum and use physical backups (XtraBackup) for fast restores, and always test restores.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:08Over-allocating memory
- 5:06Excessive indexes slow writes
- 7:39Default config pitfalls
- 9:48The myth of keeping all data in memory
- 13:01Inefficient data types waste space
- 16:18Redundant indexes dangers
- 20:35Primary keys hidden costs
- 22:02Replicas not backups
- 24:57Testing backups
- 27:07Binary vs logical backups
- 30:48Q&A
