Improving Observability, Reliability, and Security of Relational Database...
summary
LinkedIn moves hundreds-to-thousands of MySQL instances to a standardized client stack by shipping custom wrappers around open-source connectors that emit query latencies, connection metrics and pool stats and enable standardized logging and configs. Failovers use a client-side failover-to-read-write flag plus a nominated candidate cname to cut mean time to recovery from two–five minutes to 10–20 seconds. Security adds one‑click TLS and a mTLS plugin for MySQL that authenticates against certificate SAN fields. Adoption is incremental and compatible with upstream updates.
AI-written summary · about
chapters (18)
- 0:00Start
- 4:13So... What's the problem?
- 5:03The Reason
- 8:17Atomic Habits
- 9:47Challenges
- 10:08Client Tools for MySQL
- 11:09Deconstructing "The Client"
- 14:49What is a supported client?
- 16:39Why wrappers?
- 18:15Client-side metrics
- 18:41Query Metrics
- 19:32Connection
- 20:46Better logging
- 21:44Additional things to log
- 26:08Problem 2
- 29:03The limiting factor
- 34:48Making security easy
- 38:51Conclusion
related talks
46mDatabase Reliability Engineering for MySQL
41mState of MySQL Security in 2021
57mThe Evolution of a MySQL Database System
53mMySQL DevOps
54mThe Evolution of a MySQL Database, From a Single Instance to HA With Disaster Recovery
35mBest Practices for MySQL Performance
41mDemocratizing MySQL: From Cloud Managed to Kubernetes Managed
43mMonitoring MySQL at scale