Create your Own MySQL-as-a-Service that Runs Anywhere Using Kubernetes Operators
summary
DIY MySQL-as-a-Service runs on Kubernetes using Presslabs operator and Orchestrator with Percona MySQL containers, StatefulSets and a sidecar that writes backups to S3; production workload moves off AWS RDS to private cloud and runs for ~1.5 years. Automated failover promotes replicas but single-master design causes ~1–3 minute downtime on failure. Monitoring uses Prometheus; storage on OpenStack/Cinder with LVM works better than vendor controllers; current backups use mysqldump, plan snapshot backups, multi‑master and online upgrades remain unfinished.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:11About Platform Line
- 1:56Agenda
- 3:08Background
- 4:55What did you look for
- 10:21Why not Kubernetes
- 13:40Our Solution
- 16:25Components
- 17:54Cluster Architecture
- 19:37Kubernetes Containers
- 21:04Orchestrator
- 22:37Problems with both approaches
- 26:50Software stack issues
- 28:35Future plans
- 29:28Links
- 30:27Questions
related talks
41mDemocratizing MySQL: From Cloud Managed to Kubernetes Managed
35mRunning MySQL on Kubernetes [I]
23mRunning MySQL on K8s. Deployment, Scaling, Backups and advanced configuration
53mMySQL in the Cloud: DBaaS vs Kubernetes
56mDeploying MySQL on Kubernetes & Openshift
19mPrivate DBaaS: Why you Need it, How to Build It
36mFrom a Database in Container to DBaaS on Kubernetes
27mDatabase Infrastructure with Open Source Kubernetes Database Operators
