Best Practice in Design and Implementing MySQL Geographic Distributed, High Availability Solutions
summary
Claims concrete tradeoffs and concrete setups: evaluation of PXC/Galera and group replication versus async primary/secondary, measured latency effects at 400 kilometers and examples like 3 ms ping versus spikes to 37 ms and >200 ms that cause re‑syncs and stalled clusters. Shows real deployments labeled “wrong”, “almost okay”, and “best” with fiber cables versus Internet links, uses ProxySQL for failover, insists on long‑term network sampling, monitoring, backups, and that some configurations visibly fail under load.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:58Why Are We Talking About Geographic Distribu The need to have geographic distribution for...
- 3:01What is Geographic Distribution?
- 3:25What is HA and What is DR?
- 5:00The Impact of Distance on Geographic Distribution
- 5:29A Real-Life Example
- 8:48Observations
- 9:53Facts About Server 3
- 12:03TCP Encapsulation
- 13:16TCP Sliding Window
- 17:33Replicate Data is the key - Sync vs Async
- 17:53Which Implementations?: Primary - Secondary
- 19:46Which Implementations?: PXC
- 21:03Which Implementations?: PS - Group Replication Replication by standard MySQL
- 22:03Implementations by HA Level and Tight/Loose Rel. Solution distribution by HA and replication level
- 24:26Architectures: What Should NOT Be Done 2
- 26:10Architectures: What Can Be Done
- 27:19Architectures: What You Should Do
- 28:20Conclusion - a Healthy Solution
- 31:22Useful References
related talks
24mComparing the High Availability of MySQL with Modern Distributed SQL Databases
52mBest Practices for MySQL High Availability in 2022
35mMySQL HA, Recovery and Load Balancing
27mDistributed MySQL Architectures: Past, Present and Future
49mMySQL Replication — Advanced Features
21mHow Safe is Asynchronous Master-Master Setup?
46mDatabase Reliability Engineering for MySQL
53mMySQL DevOps