Confessions of a DBA: worst and best things I've done in production
summary
Runs 3,000 customer databases in Amazon and shows concrete fixes: converts a heavy write table with ALTER TABLE ... ENGINE=BLACKHOLE to stop replication lag, discovers a catastrophic mv vs cp sparse files bug when moving InnoDB tables across partitions, and rebuilds a corrupt InnoDB unique index by sharding data into ~14,165 slices with bash scripts and remerging. Calls out useless MySQL Tuner output, uses Percona tools like pt-query-digest, dumps stats to statsd, and prefers live experiments (turning off query cache) over heuristics.
AI-written summary · about
chapters (12)
- 0:00Intro
- 3:09Timeline
- 9:15Data Migration
- 14:33RAID
- 18:36My Sequel
- 21:11Bad Pearl
- 22:57Bad Node
- 25:03Big Registry
- 27:34One bad table
- 31:17Benchmarking
- 33:46Query Cache
- 37:00Open Source
related talks
28mBecoming a DBA for 5 minutes a day, the most important things to know for sysadmins and developers
37mNavigating Managed Databases for Over a Decade as a Non-DBA
53mMySQL DevOps
44mBridging the Gap: The Vital Role of DBAs while contributing to the DevOps Culture
01mAre you an "accidental" DBA?
01h 57mPercona Community Live Conference - Day 2 - Open Source Databases Conference
12mAdvanced database topics and Database 101 for Developers
47mFun with BIG Data with various databases