Running Vitess on Kubernetes at Massive Scale: JD.com Case Study
summary
Vitess runs on Kubernetes with VTGate, VTTablet and VTCTLD backed by etcd and MySQL; JD.com deploys it across Kubernetes cells in eight data centers and reports 1,000+ keyspaces, ~10,000 shards and ~200+ TB of data with the largest keyspace at 128 shards, serving massive QPS. Resharding proceeds by consuming MySQL binary logs, unplanned reparent uses orchestrator, replicas use semi‑sync, and JD implements single‑container MySQL+tablet, local PVs, custom DPS and VLog collection; reshard isolation, VSchema distribution and auto‑balancing remain active engineering tasks.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:22Jitens background
- 4:10Vitess architecture
- 10:30Vitess Gate
- 11:44Control Plane
- 12:58HCD
- 14:17Tablets
- 15:50High Availability
- 16:42About JDcom
- 18:04Before Vitess
- 19:00Vitess Advantages
- 20:18Vitess Limitations
- 21:20JDcom Vitess
- 25:12Vitess Steam
- 27:20Vitess Ecosystem
- 31:53Metadata
- 33:25Orchestration
- 34:22Future plans
related talks
35mTwo Years with Vitess: How JD.com Runs the World's Largest Vitess
35mTwo Years with Vitess: How JD.com Runs the World's Largest Vitess
31mVitess: Running MySQL with confidence on Kubernetes
43mVitess and a Kubernetes Operator
44mIntro: Vitess
12mKeynote: Keeping Your Data in Kubernetes with Vitess
38mIntro: Vitess
34mInside Vitess: Building Resilient, Distributed Databases