Scaling Microservices Beyond a Single Cluster w/ Kubernetes
summary
Concur runs multiple Kubernetes clusters per location and uses external F5 load balancers plus an automated Go-based watcher container that watches the Kubernetes API and calls an internal LB API to register services and nodes, keeping services online if a whole cluster fails. CI/infra includes Terraform, CoreOS, flannel, Artifactory, Jenkins/Circle/CICD, and ELK moved outside clusters. A custom Skipper deployment API handles multi-cluster and blue-green Kubernetes clusters upgrades; it works in production but complicates stateful apps and troubleshooting.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:12Agenda
- 3:12Tools
- 6:54Clusters
- 21:19Demo
- 31:29Artifact
- 36:56Benefits
- 38:54Challenges
- 42:44Other Projects
related talks
44mScaling Microservices with Kubernetes
46mWhy Kubernetes is Not Enough
31mHere Be Services: Beyond the Cluster Boundary with Multicluster Servi...
49mDeveloping microservices applications on the Kubernetes platform
53mManaging Containers at Scale with CoreOS and Kubernetes
35mTo Infinite Scale and Beyond: Operating Kubernetes Past the Steady State
53mNext Generation Microservices
50mMulti-Cluster Kubernetes and Service Mesh Patterns