Install CloudGrange — On-Premises

On-premises CloudGrange installs a local control plane that discovers and manages your existing clusters and VMs without requiring an Azure-hosted control plane.

CloudGrange is in preview, not GA or a certified production release. Every install page downloads the latest release. See current product and release status.

What gets installed

CloudGrange runs on Kubernetes and is deployed by Helm. Every on-premises installation path ends at the same Helm chart on a Kubernetes cluster — the paths differ only in how much of the stack beneath the chart they build for you.

Component Kubernetes workload
API (cloudgrange-api) Deployment — the platform REST API
Portal (cloudgrange-portal) Deployment — the web UI
Relay Deployment + Service — per-site gateway routing jobs to host agents
PostgreSQL StatefulSet with a PersistentVolumeClaim — database and encrypted secrets store
Keycloak Deployment — local identity provider
Observability OpenTelemetry collector, Prometheus, Loki, Grafana
cert-manager Separate Helm release — issues the in-cluster TLS certificate

An Ingress publishes 443 (portal/API/identity) and 80 (redirect). The relay's agent API is exposed on 8443 through its own Service. Everything else stays on the cluster network.

Where no Kubernetes cluster exists, the installers deploy K3s — a single-binary, CNCF-conformant Kubernetes distribution that runs comfortably on one VM and can later grow to a multi-node HA control plane without re-platforming.

The Docker Compose stack used by earlier previews is legacy. It remains available for one release cycle via --engine compose and will be retired.

The four installation paths

All four deploy the same chart. Pick by what you already have:

Path You provide The installer builds
Appliance (VHDX) A Windows/Hyper-V host Import a prepared VHDX — Linux, K3s and the chart are already baked in. Fully offline.
Windows-orchestrated A Windows/Hyper-V host Creates the Linux VM, installs K3s inside it, deploys the chart. Also available offline/bundled.
Native Linux An Ubuntu/Debian server Installs K3s on that server, deploys the chart.
Helm Any Kubernetes cluster Nothing — installs the chart only. microk8s, kind, RKE2, EKS, AKS, GKE.

If you already run Kubernetes, use the Helm path. It is the least invasive and installs nothing on your hosts.

Start installing

If your control-plane host is… Use this guide
A Kubernetes cluster you already operate Install with Helm
An existing Ubuntu or Debian server Install on Native Linux
An internet-connected Windows/Hyper-V host Install in Online mode
A Windows/Hyper-V host that must install offline Install in Bundled mode
A Windows/Hyper-V host where you want to import a prepared VHDX Import the Appliance

Each guide contains the current release download URL, checksum procedure, exact command, expected result, and first-run steps. There is no PaaS installation procedure because PaaS has not been released.

Published SHA-256 manifests check integrity only; preview artifacts are not signed.

Sizing

Path vCPU RAM Disk
Windows script VM (tested configuration) 4 8 GB static 30 GB virtual disk
VHDX appliance VM 2 4 GB startup, 2–8 GB dynamic The imported VHDX

CloudGrange's workloads request 870m CPU and 1.75 GiB memory, and their memory limits total about 5.4 GiB. For every path's requirements, see Prerequisites.

The database, Prometheus and Loki volumes grow with managed-host count and retention.

Next steps

  1. Review the detailed on-premises install guide.
  2. Install the Relay at each site.
  3. Install the Agent on each Hyper-V host.

See current product and release status.