How to install the Radius control plane

There are three ways to install the Radius control plane. They install the same components on the Kubernetes cluster; they differ in how much of the surrounding setup they automate and how much control they give you over the installation.

• rad install performs the base installation and is oriented toward platform engineers who are setting up a centralized Radius control plane for developers to connect to. rad install accepts --set flags to override individual Helm chart values (listed below).
• rad initialize does everything rad install does plus configures the local workstation with a Workspace and Bicep configuration file. It is oriented toward local development, where you install Radius and build an application on the same machine.
• helm upgrade radius installs only the control plane. It does not create Resource Types, Resource Groups, or Environments.

The following table summarizes the differences:

Each method is described in detail below.

Using rad initialize

rad initialize is the fastest way to get started. It installs the control plane on the current Kubernetes cluster and configures the local machine in one step:

rad initialize --preview

Customizing with rad initialize --full

By default, rad initialize uses opinionated defaults: it installs into the current Kubernetes context, creates an Environment named default that deploys applications into the default namespace, and does not configure any cloud providers.

To choose these settings yourself, run rad initialize with the --full flag:

rad initialize --full --preview

In full mode, rad initialize interactively prompts you for all available configuration options, including:

• The Kubernetes cluster to install onto
• The Environment name
• The namespace that applications are deployed into
• Azure and AWS cloud providers and their credentials

Use --full when the defaults do not match your setup, for example when you want a non-default Environment or namespace, or when you need to configure cloud providers during initialization.

Using rad install

Use rad install kubernetes to install the control plane. It installs the control plane, creates the default Resource Types, and creates a default Resource Group and Environment, but it does not create any local files. Optionally use the --set flag to customize the installation with Helm chart options:

# Install Radius
rad install kubernetes --preview

# Install Radius with tracing and a public endpoint override
rad install kubernetes --preview --set global.zipkin.url=http://jaeger-collector.radius-monitoring.svc.cluster.local:9411/api/v2/spans,rp.publicEndpointOverride=localhost:8081

Use your own root certificate authority certificate

Many enterprises leverage intermediate root certificate authorities (CAs) to enhance security and control over outgoing traffic, particularly when using a firewall or proxy. In this setup, when Radius attempts to connect to an external endpoint such as Azure or AWS, traffic may be blocked by the firewall. Optionally use --set-file when installing Radius to inject your root CA certificate into Radius:

rad install kubernetes --preview --set-file global.rootCA.cert=/etc/ssl/your-root-ca.crt

Using Helm

For full control over the installation, install Radius directly with Helm. The Radius chart is published as an OCI artifact to GitHub Container Registry (GHCR), so there is no Helm repository to add.

Install the chart directly from its OCI registry:

helm upgrade radius oci://ghcr.io/radius-project/helm-chart --install --create-namespace --namespace radius-system --version 0.26.0 --wait --timeout 15m0s

To install a different version, change the --version flag. Inspect the chart before installing with:

helm show chart oci://ghcr.io/radius-project/helm-chart --version 0.26.0

Customizing the installation with Helm chart options

Whether you install with rad install kubernetes --set or with Helm directly, the following chart options are available:

Next steps

Once the Radius control plane has been installed, Radius can be used to deploy applications to the Kubernetes cluster. In order to deploy resources to AWS or Azure, configure cloud providers.