Skip to main content

K3s Cluster Topology

The platform runs on a 3-node K3s cluster — a lightweight, certified Kubernetes distribution. The cluster consists of 1 server (control plane) node and 2 worker nodes.

Node Architecture​

Scheduling & Availability​

The cluster uses several mechanisms to distribute workloads and maintain availability across the 3 nodes:

  • Soft node affinity: Applications use preferredDuringSchedulingIgnoredDuringExecution instead of hard nodeSelector, allowing the scheduler flexibility when nodes are under pressure
  • Topology spread constraints: Multi-replica deployments spread pods across nodes using topologySpreadConstraints with maxSkew: 1
  • PriorityClasses: Cluster-wide priority classes ensure critical workloads (databases, ingress, monitoring) are scheduled before lower-priority pods during resource contention
  • Resource quotas: Namespace-level resource quotas prevent any single namespace from consuming all cluster resources

Namespace Layout​

Application Deployment Pattern​

Each application follows a consistent pattern:

  • Client deployment: Nginx serving the built frontend (React/Storybook)
  • Server deployment (where applicable): Node.js Express/Apollo API server
  • Service: ClusterIP service for internal routing
  • Ingress: Standard Kubernetes Ingress with ingressClassName: traefik
  • HPA: Horizontal Pod Autoscaler for auto-scaling
Interactive Components

Explore extracted UI components from each application in the Storybook Showcase. See the Applications Overview for per-app story links.

Example: Bookmarked​