Use Azure RBAC on Azure Arc-enabled Kubernetes clusters

Kubernetes ClusterRoleBinding and RoleBinding object types help to define authorization in Kubernetes natively. By using this feature, you can use Microsoft Entra ID and role assignments in Azure to control authorization checks on the cluster. Azure role assignments let you granularly control which users can read, write, and delete Kubernetes objects such as deployment, pod, and service.

For a conceptual overview of this feature, see Azure RBAC on Azure Arc-enabled Kubernetes.

Prerequisites

• Install or upgrade the Azure CLI to the latest version.

• Install the latest version of connectedk8s Azure CLI extension:

  az extension add --name connectedk8s
  

  If the connectedk8s extension is already installed, you can update it to the latest version by using the following command:

  az extension update --name connectedk8s
  

• Connect an existing Azure Arc-enabled Kubernetes cluster:

  • If you haven't connected a cluster yet, use our quickstart.
  • Upgrade your agents to the latest version.

Enable Azure RBAC on the cluster

1. Get the cluster MSI identity by running the following command:

   az connectedk8s show -g <resource-group> -n <connected-cluster-name>
   

2. Get the ID (identity.principalId) from the output and run the following command to assign the Connected Cluster Managed Identity CheckAccess Reader role to the cluster MSI:

   az role assignment create --role "Connected Cluster Managed Identity CheckAccess Reader" --assignee "<Cluster MSI ID>" --scope <cluster ARM ID>
   

3. Enable Azure role-based access control (RBAC) on your Azure Arc-enabled Kubernetes cluster by running the following command:

   az connectedk8s enable-features -n <clusterName> -g <resourceGroupName> --features azure-rbac
   

   Note

   Before you run the preceding command, ensure that the kubeconfig file on the machine is pointing to the cluster on which you'll enable the Azure RBAC feature.

   Use --skip-azure-rbac-list with the preceding command for a comma-separated list of usernames, emails, and OpenID connections undergoing authorization checks by using Kubernetes native ClusterRoleBinding and RoleBinding objects instead of Azure RBAC.

Generic cluster where no reconciler is running on the apiserver specification

1. SSH into every master node of the cluster and take the following steps:

   If your kube-apiserver is a static pod:

   1. The azure-arc-guard-manifests secret in the kube-system namespace contains two files: guard-authn-webhook.yaml and guard-authz-webhook.yaml. Copy these files to the /etc/guard directory of the node.

      sudo mkdir -p /etc/guard
      kubectl get secrets azure-arc-guard-manifests -n kube-system -o json | jq -r '.data."guard-authn-webhook.yaml"' | base64 -d > /etc/guard/guard-authn-webhook.yaml
      kubectl get secrets azure-arc-guard-manifests -n kube-system -o json | jq -r '.data."guard-authz-webhook.yaml"' | base64 -d > /etc/guard/guard-authz-webhook.yaml
      

   2. Open the apiserver manifest in edit mode:

      sudo vi /etc/kubernetes/manifests/kube-apiserver.yaml
      

   3. Add the following specification under volumes:

      - name: azure-rbac
          hostPath:
          path: /etc/guard
          type: Directory
      

   4. Add the following specification under volumeMounts:

      - mountPath: /etc/guard
          name: azure-rbac
          readOnly: true
      

   If your kube-apiserver is a not a static pod:

   1. Open the apiserver manifest in edit mode:

      sudo vi /etc/kubernetes/manifests/kube-apiserver.yaml
      

   2. Add the following specification under volumes:

      - name: azure-rbac
          secret:
          secretName: azure-arc-guard-manifests
      

   3. Add the following specification under volumeMounts:

      - mountPath: /etc/guard
          name: azure-rbac
          readOnly: true
      

2. Add the following apiserver arguments:

   - --authentication-token-webhook-config-file=/etc/guard/guard-authn-webhook.yaml
   - --authentication-token-webhook-cache-ttl=5m0s
   - --authorization-webhook-cache-authorized-ttl=5m0s
   - --authorization-webhook-config-file=/etc/guard/guard-authz-webhook.yaml
   - --authorization-webhook-version=v1
   - --authorization-mode=Node,RBAC,Webhook
   

   If the Kubernetes cluster is version 1.19.0 or later, you also need to set the following apiserver argument:

   - --authentication-token-webhook-version=v1
   

3. Save and close the editor to update the apiserver pod.

Cluster created by using Cluster API

1. Copy the guard secret that contains authentication and authorization webhook configuration files from the workload cluster onto your machine:

   kubectl get secret azure-arc-guard-manifests -n kube-system -o yaml > azure-arc-guard-manifests.yaml
   

2. Change the namespace field in the azure-arc-guard-manifests.yaml file to the namespace within the management cluster where you're applying the custom resources for creation of workload clusters.

3. Apply this manifest:

   kubectl apply -f azure-arc-guard-manifests.yaml
   

4. Edit the KubeadmControlPlane object by running kubectl edit kcp <clustername>-control-plane:

   1. Add the following snippet under files:

      - contentFrom:
          secret:
            key: guard-authn-webhook.yaml
            name: azure-arc-guard-manifests
        owner: root:root
        path: /etc/kubernetes/guard-authn-webhook.yaml
        permissions: "0644"
      - contentFrom:
          secret:
            key: guard-authz-webhook.yaml
            name: azure-arc-guard-manifests
        owner: root:root
        path: /etc/kubernetes/guard-authz-webhook.yaml
        permissions: "0644"
      

   2. Add the following snippet under apiServer > extraVolumes:

      - hostPath: /etc/kubernetes/guard-authn-webhook.yaml
          mountPath: /etc/guard/guard-authn-webhook.yaml
          name: guard-authn
          readOnly: true
      - hostPath: /etc/kubernetes/guard-authz-webhook.yaml
          mountPath: /etc/guard/guard-authz-webhook.yaml
          name: guard-authz
          readOnly: true
      

   3. Add the following snippet under apiServer > extraArgs:

      authentication-token-webhook-cache-ttl: 5m0s
      authentication-token-webhook-config-file: /etc/guard/guard-authn-webhook.yaml
      authentication-token-webhook-version: v1
      authorization-mode: Node,RBAC,Webhook
      authorization-webhook-cache-authorized-ttl: 5m0s
      authorization-webhook-config-file: /etc/guard/guard-authz-webhook.yaml
      authorization-webhook-version: v1
      

   4. Save and close to update the KubeadmControlPlane object. Wait for these changes to appear on the workload cluster.

Create role assignments for users to access the cluster

Owners of the Azure Arc-enabled Kubernetes resource can use either built-in roles or custom roles to grant other users access to the Kubernetes cluster.

Built-in roles

Role	Description
Azure Arc Kubernetes Viewer	Allows read-only access to see most objects in a namespace. This role doesn't allow viewing secrets, because read permission on secrets would enable access to ServiceAccount credentials in the namespace. These credentials would in turn allow API access through that ServiceAccount value (a form of privilege escalation).
Azure Arc Kubernetes Writer	Allows read/write access to most objects in a namespace. This role doesn't allow viewing or modifying roles or role bindings. However, this role allows accessing secrets and running pods as any ServiceAccount value in the namespace, so it can be used to gain the API access levels of any ServiceAccount value in the namespace.
Azure Arc Kubernetes Admin	Allows admin access. It's intended to be granted within a namespace through RoleBinding. If you use it in RoleBinding, it allows read/write access to most resources in a namespace, including the ability to create roles and role bindings within the namespace. This role doesn't allow write access to resource quota or to the namespace itself.
Azure Arc Kubernetes Cluster Admin	Allows superuser access to execute any action on any resource. When you use it in ClusterRoleBinding, it gives full control over every resource in the cluster and in all namespaces. When you use it in RoleBinding, it gives full control over every resource in the role binding's namespace, including the namespace itself.

You can create role assignments scoped to the Azure Arc-enabled Kubernetes cluster in the Azure portal on the Access Control (IAM) pane of the cluster resource. You can also use the following Azure CLI commands:

az role assignment create --role "Azure Arc Kubernetes Cluster Admin" --assignee <AZURE-AD-ENTITY-ID> --scope $ARM_ID

In those commands, AZURE-AD-ENTITY-ID can be a username (for example, testuser@mytenant.partner.onmschina.cn) or even the appId value of a service principal.

Here's another example of creating a role assignment scoped to a specific namespace within the cluster:

az role assignment create --role "Azure Arc Kubernetes Viewer" --assignee <AZURE-AD-ENTITY-ID> --scope $ARM_ID/namespaces/<namespace-name>

Custom roles

You can choose to create your own role definition for use in role assignments.

Walk through the following example of a role definition that allows a user to only read deployments. For more information, see the full list of data actions that you can use to construct a role definition.

Copy the following JSON object into a file called custom-role.json. Replace the <subscription-id> placeholder with the actual subscription ID. The custom role uses one of the data actions and lets you view all deployments in the scope (cluster or namespace) where the role assignment is created.

{
    "Name": "Arc Deployment Viewer",
    "Description": "Lets you view all deployments in cluster/namespace.",
    "Actions": [],
    "NotActions": [],
    "DataActions": [
        "Microsoft.Kubernetes/connectedClusters/apps/deployments/read"
    ],
    "NotDataActions": [],
    "assignableScopes": [
        "/subscriptions/<subscription-id>"
    ]
}

1. Create the role definition by running the following command from the folder where you saved custom-role.json:

   az role definition create --role-definition @custom-role.json
   

2. Create a role assignment by using this custom role definition:

   az role assignment create --role "Arc Deployment Viewer" --assignee <AZURE-AD-ENTITY-ID> --scope $ARM_ID/namespaces/<namespace-name>
   

Configure kubectl with user credentials

There are two ways to get the kubeconfig file that you need to access the cluster:

• You use the Cluster Connect feature (az connectedk8s proxy) of the Azure Arc-enabled Kubernetes cluster.
• The cluster admin shares the kubeconfig file with every other user.

Use cluster connect

Run the following command to start the proxy process:

az connectedk8s proxy -n <clusterName> -g <resourceGroupName>

After the proxy process is running, you can open another tab in your console to start sending your requests to the cluster.

Use a shared kubeconfig file

Using a shared kubeconfig requires slightly different steps depending on your Kubernetes version.

Send requests to the cluster

1. Run any kubectl command. For example:

   • kubectl get nodes
   • kubectl get pods

2. After you're prompted for browser-based authentication, copy the device login URL (https://microsoft.com/deviceloginchina) and open it in your web browser.

3. Enter the code printed on your console. Copy and paste the code on your terminal into the prompt for device authentication input.

4. Enter the username (testuser@mytenant.partner.onmschina.cn) and the associated password.

5. If you see an error message like this, it means you're unauthorized to access the requested resource:

   Error from server (Forbidden): nodes is forbidden: User "testuser@mytenant.partner.onmschina.cn" cannot list resource "nodes" in API group "" at the cluster scope: User doesn't have access to the resource in Azure. Update role assignment to allow access.
   

   An administrator needs to create a new role assignment that authorizes this user to have access on the resource.

Use Conditional Access with Microsoft Entra ID

When you're integrating Microsoft Entra ID with your Azure Arc-enabled Kubernetes cluster, you can also use Conditional Access to control access to your cluster.

To create an example Conditional Access policy to use with the cluster:

1. At the top of the Azure portal, search for and select Microsoft Entra ID.

2. On the menu for Microsoft Entra ID on the left side, select Enterprise applications.

3. On the menu for enterprise applications on the left side, select Conditional Access.

4. On the menu for Conditional Access on the left side, select Policies > New policy.

   

5. Enter a name for the policy, such as arc-k8s-policy.

6. Select Users and groups. Under Include, choose Select users and groups. Then choose the users and groups where you want to apply the policy. For this example, choose the same Microsoft Entra group that has administrative access to your cluster.

   

7. Select Cloud apps or actions. Under Include, choose Select apps. Then search for and select the server application that you created earlier.

   

8. Under Access controls, select Grant. Select Grant access > Require device to be marked as compliant.

   

9. Under Enable policy, select On > Create.

   

Access the cluster again. For example, run the kubectl get nodes command to view nodes in the cluster:

kubectl get nodes

Follow the instructions to sign in again. An error message states that you're successfully logged in, but your admin requires the device that's requesting access to be managed by Microsoft Entra ID in order to access the resource. Follow these steps:

1. In the Azure portal, go to Microsoft Entra ID.

2. Select Enterprise applications. Then under Activity, select Sign-ins.

3. An entry at the top shows Failed for Status and Success for Conditional Access. Select the entry, and then select Conditional Access in Details. Notice that your Conditional Access policy is listed.

   

Configure just-in-time cluster access with Microsoft Entra ID

Another option for cluster access control is to use Privileged Identity Management (PIM) for just-in-time requests.

To configure just-in-time access requests for your cluster, complete the following steps:

1. At the top of the Azure portal, search for and select Microsoft Entra ID.

2. Take note of the tenant ID. For the rest of these instructions, we'll refer to that ID as <tenant-id>.

   

3. On the menu for Microsoft Entra ID on the left side, under Manage, select Groups > New group.

4. Make sure that Security is selected for Group type. Enter a group name, such as myJITGroup. Under Microsoft Entra roles can be assigned to this group (Preview), select Yes. Finally, select Create.

   

5. You're brought back to the Groups page. Select your newly created group and take note of the object ID. For the rest of these instructions, we'll refer to this ID as <object-id>.

   

6. Back in the Azure portal, on the menu for Activity on the left side, select Privileged Access (Preview). Then select Enable Privileged Access.

   

7. Select Add assignments to begin granting access.

   

8. Select a role of Member, and select the users and groups to whom you want to grant cluster access. A group admin can modify these assignments at any time. When you're ready to move on, select Next.

   

9. Choose an assignment type of Active, choose the desired duration, and provide a justification. When you're ready to proceed, select Assign. For more on assignment types, see Assign eligibility for a privileged access group (preview) in Privileged Identity Management.

   

After you've made the assignments, verify that just-in-time access is working by accessing the cluster. For example, use the kubectl get nodes command to view nodes in the cluster:

kubectl get nodes

Note the authentication requirement and follow the steps to authenticate. If authentication is successful, you should see output similar to this:

To sign in, use a web browser to open the page https://microsoft.com/deviceloginchina and enter the code AAAAAAAAA to authenticate.

NAME      STATUS   ROLES    AGE      VERSION
node-1    Ready    agent    6m36s    v1.18.14
node-2    Ready    agent    6m42s    v1.18.14
node-3    Ready    agent    6m33s    v1.18.14

Next steps

• Securely connect to the cluster by using Cluster Connect.
• Read about the architecture of Azure RBAC on Arc-enabled Kubernetes.