Database Reliability Engineering for MySQL
summary
Describes a concrete MySQL reliability stack: Orchestrator with Raft (requires three nodes) automates primary failover, uses holistic detection and an anti‑flap one‑hour grace; ProxySQL core+satellite cluster provides service discovery and query routing with host groups and query rules; observability via PMM (VictoriaMetrics time series + ClickHouse query analytics) with exporters and alerting; backups via Percona XtraBackup or disk snapshots and binary‑log shipping for point‑in‑time recovery, noting InnoDB preference and RTO/RPO tradeoffs.
AI-written summary · about
chapters (22)
- 0:00Intro
- 2:03Availability vs High Availability
- 4:40High Availability for MySQL
- 5:11MySQL - Single point of failure
- 6:20Orchestrator: Discovery
- 7:11Orchestrator: Visualization
- 8:24Orchestrator: Refactoring
- 9:49Orchestrator: Recovery
- 12:40MySQL HA-back to our example
- 14:20Orchestrator/raft
- 16:52Orchestrator shared db vs raft
- 17:54What is service discovery?
- 20:04ProxySQL: What?
- 22:16ProxySQL: Terminology
- 24:02ProxySQL: Basic design (2)
- 26:32What is observability
- 28:03Percona Monitoring and Management (PMM)
- 28:15What is PMM?
- 30:41PMM: Components
- 36:25What is Disaster Recovery (DR)
- 39:20Backups - Percona Xtrabackup
- 41:18Point-in-time recovery
related talks
21mDatabase Reliability Engineering
40mImproving Observability, Reliability, and Security of Relational Database...
52mBest Practices for MySQL High Availability in 2022
54mThe Evolution of a MySQL Database, From a Single Instance to HA With Disaster Recovery
57mThe Evolution of a MySQL Database System
43mMySQL Failover for Asynchronous replication with Orchestrator, ProxySQL and Consul
32mBest Practice in Design and Implementing MySQL Geographic Distributed, High Availability Solutions
05mMySQL failover with orchestrator and ProxySQL