Skip to content

Cluster architecture

Context

When you create a cluster you can download kubeconfig to access your cluster directly. Some of the middleware that is running on the cluster (Grafana, Kubernetes dashboard) is directly reachable from console UI through the single sign-on gateway.

You can notice on the diagram below that there is a bi-direction link between your cluster and CAST AI platform. Not only the platform connects to your cloud infrastructure or the cluster itself; CAST AI also relies on the cluster to "call back" and inform about certain events:

  • Cluster control plane nodes actions with provisioning engine, e.g. when to join the cluster;
  • Nodes inform about operations being completed, like finishing joining the cluster;
  • Relevant cloud events get propagated to provisioning engine & autoscaler, for example, "spot instance is being terminated by cloud provider";

Your app users do not interact with CAST AI in any way. You own your kubernetes cluster infrastructure 100%, including any ingress infrastructure to reach your cluster workloads.

Below diagram highlights primary groups of components that define a relationship between CAST AI platform and your cluster: