Building a Database as a Service on Kubernetes
Talkat KubeCon + CloudNativeCon North America 2019 · on Tuesday, 2019-11-19· at 14:25· Room 30ABCDE - San Diego Convention Center, Upper Level·
summary
PlanetScale describes a production Database-as-a-Service on Kubernetes that uses Vitess over MySQL, a rewritten Vitess operator to provision clusters, etcd per tenant for topology metadata and PVC-backed statefulsets for durability, and Prometheus + Grafana for monitoring. Deployment takes 2–4 minutes per database, HA uses three instances, and VTGate pools behind per-tenant load balancers route MySQL connections. Remaining work: BYOK and true multi‑cloud masters/replicas.
AI-written summary · about
chapters (20)
- 0:00Intro
- 0:34Who are we
- 0:58Agenda
- 1:36Mutable Infrastructure
- 5:45Minimum Viable Product
- 6:48Why Open Source
- 8:09Why Kubernetes
- 9:10The Test
- 11:23Operators
- 12:33SCD Operator
- 13:47Prometheus
- 15:47Graph Fauna
- 16:35Thin Occlusion Proxy
- 17:08Architecture Diagram
- 21:19Was it hard
- 22:47Whats next
- 24:11Questions
- 24:46The Operator
- 27:14Expert
- 33:37Load Balancer
related talks
31mVitess: Running MySQL with confidence on Kubernetes
05mRun Your Data Layer In K8s With Vitess
38mIntro: Vitess
36mHorizontal Scaling with Vitess
43mVitess and a Kubernetes Operator
34mInside Vitess: Building Resilient, Distributed Databases
35mExploring High Availability in Kubernetes with Vitess
40mVitess: Intro and New Features