Skip to content

Installation

DC Suite runs on Kubernetes. This page covers the shape of an install; adapt the specifics to your environment.

Prerequisites

  • A Kubernetes cluster to host the control plane.
  • PostgreSQL (managed or in-cluster) for the system of record.
  • A workflow engine endpoint for durable operations.
  • A reachable fleet manager managing your bare-metal pool.
  • An identity provider for SSO.
  • (Optional) metrics and logs backends for observability.
  • An ingress controller and wildcard DNS for the console and for the per-cluster public URLs DC Suite mints.
  • A container registry the cluster can pull the DC Suite images from.

Components to deploy

Workload Notes
API server The control plane. Needs DB, workflow, fleet-manager, SSO config.
Web console Static SPA served behind the same domain.
Provisioning runner Runs per-cluster infrastructure jobs.
PostgreSQL If not using a managed database.

Install with Helm

A Helm chart packages the workloads and their configuration. At a high level:

helm upgrade --install dc-suite ./deploy/helm/dc-suite \
  --namespace dc-suite --create-namespace \
  --values my-values.yaml

Your values.yaml supplies environment (prod), the backend endpoints, SSO settings, and image tags. See Configuration for the keys.

Ingress and DNS

DC Suite needs two kinds of routable names:

  1. The console hostname (e.g. dc-suite.example.com).
  2. A wildcard for per-cluster public URLs (e.g. *.clusters.example.com), so each exposed cluster port gets a stable HTTPS hostname with a certificate.

Point the wildcard at your ingress controller and give it a certificate resolver (for example ACME) so new hostnames get certificates automatically.

Bring machines into the pool

Deploying the control plane doesn't give you capacity — you still need to onboard physical machines through the fleet manager. See Bare-Metal Onboarding.

Verify

  • The console loads and redirects to SSO.
  • GET /healthz returns OK.
  • GET /v1/profiles lists your hardware profiles with availability once machines are onboarded.
  • A test cluster reaches READY and can be deleted cleanly.

Next: Configuration.