What We’ve Learned Building a Multi-Region DBaaS on Kubernetes
summary
Cockroach Cloud runs a global DBaaS on Kubernetes, with 70 production clusters since November, using one k8s cluster per customer and statefulsets plus persistent volumes for data durability. Multi‑region node‑to‑node networking is solved with DNS chaining: exposing kube DNS via service type load balancers and proxying lookups across regions. Automation calls cloud APIs for AWS and GKE node pools. Notable bug: PVs land in wrong zones due to wait for first consumer unset and a race when scaling.
AI-written summary · about
chapters (7)
- 0:00Start
- 3:41What is Cockroach DB
- 4:17CockroachDB Architecture
- 10:21Rolling updates
- 11:49Bug with scaling
- 13:50Kes networking
- 14:21CoockroachCloud multi-region networking
related talks
14mHow CockroachDB can enable Multi-Region Deployments and Data Locality 
22mRunning distributed stateful system on Kubernetes - Beyond Statefulsets
48mBuilding mission-critical applications with CockroachDB
45mMulti-Kubernetes, Multi-Region — Help!
26mTaking Your Database Beyond the Border of a Single Kubernetes Cluster
31mFuture-proofing operational data at scale
26mChallenges of Managing Stateful Applications in Kubernetes
29mHow City Storage Systems designed its Data Infrastructure to minimize toil