Skip to content

Clusters and Nodes

On this page you will learn about what a Cluster and a Node are and how they fit into Kelvin.

What is a Cluster?

A Cluster is a Kubernetes cluster. It can be created by you or by Kelvin — this decision is made during initial setup.

Not familiar with Kubernetes? Leave it to Kelvin. You only need to provide a machine with a basic Linux server OS (like Ubuntu Server) and Kelvin handles the rest.

Clusters and Nodes

A Cluster is made up of one or more Nodes. Each Node is a worker machine controlled by the Cluster's Control Plane.

  • One Node per physical machine is recommended.
  • The Control Plane decides which Nodes run which workloads.
  • Nodes can be added at any time to scale capacity without affecting operations.
  • All Nodes must be on the same network.

The main limit on what you can run at the edge is hardware. Plan your infrastructure accordingly — Machine Learning workloads and multi-asset Connectors need more CPU and RAM. Each App has a reserved allocation, so more Apps require more RAM.

What Runs on a Cluster?

Each Cluster runs three types of workloads: Connectors, Kelvin SmartApps™, and Docker Apps.

The Kelvin API Client (Python) lets SmartApps™ connect into all Kelvin API features:

  • Receive or send Asset data
  • Process time series data chunks
  • Write custom calculated Data Streams to the cloud database
  • Create Control Changes and Recommendations
  • Manage Applications, Assets, Data Streams, Clusters, and Nodes

Docker Apps use the same API if they need to interact with Kelvin data — but it is not required. Any Docker-compatible container can run at the edge without any Kelvin integration.

Scaling

Scaling means adding or removing Nodes from a Cluster to match your processing needs. This happens on demand without affecting current operations — Kelvin and Kubernetes automatically move Connections and Applications to rebalance the Cluster.

Kubernetes Cluster Options

Clusters can be installed on a bare metal OS machine or on a pre-existing Kubernetes cluster.

Any Kubernetes distribution that conforms to the CNCF Kubernetes Conformance Program is supported. Common options:

  • Self-hosted: k3s, kubernetes, minikube
  • Amazon: EKS
  • Azure: AKS
  • Google: GKE

Bare Metal OS Machine

Install Ubuntu Server on a machine and run the Kelvin install script. It downloads and configures a k3s Kubernetes cluster automatically.

In this setup, each machine is a separate Cluster with one Node. To run multiple Nodes in a single Cluster, use your own Kubernetes setup.

Bring Your Own Kubernetes Cluster

Download a YAML configuration file from Kelvin and run a kubectl command to install the Kelvin infrastructure into your existing cluster.

You manage the Kubernetes cluster. Kelvin provides analytics and upgrade tooling for the Kelvin infrastructure running inside it.

Cluster Security

Kelvin installs custom pods on each Node. If Kelvin manages the Kubernetes cluster, it also manages the security of those pods — you control when upgrades happen.

You are still responsible for OS security: keep software patches up to date and maintain an active firewall.

If you manage your own cluster (K3s, AKS, EKS, etc.), follow your distribution's security recommendations.

Firewalls

  • Kelvin-managed cluster (k3s): See k3s networking requirements.
  • Self-managed cluster: Consult your Kubernetes provider's documentation.

Cluster Communications

All traffic between a Cluster and the Kelvin Cloud is encrypted with SSL.