AAD Pod Identity enables Kubernetes applications to access cloud resources securely with Azure Active Directory (AAD).
Using Kubernetes primitives, administrators configure identities and bindings to match pods. Then without any code modifications, your containerized applications can leverage any resource in the cloud that depends on AAD as an identity provider.
You will need a Kubernetes cluster running on Azure, either managed by AKS or provisioned with AKS Engine.
AAD Pod Identity consists of the Managed Identity Controller (MIC) deployment, the Node Managed Identity (NMI) daemon set, and several standard and custom resources. For more information, see Components.
Run this command to create the aad-pod-identity
deployment on an RBAC-enabled cluster:
kubectl apply -f https://raw.githubusercontent.com/Azure/aad-pod-identity/master/deploy/infra/deployment-rbac.yaml
Or run this command to deploy to a non-RBAC cluster:
kubectl apply -f https://raw.githubusercontent.com/Azure/aad-pod-identity/master/deploy/infra/deployment.yaml
Run this Azure CLI command, and take note of the clientId
and id
values it returns:
az identity create -g <resourcegroup> -n <name> -o json
Here is an example of the output:
$ az identity create -g myresourcegroup -n myidentity -o json
{
"clientId": "00000000-0000-0000-0000-000000000000",
"clientSecretUrl": "https://control-eastus.identity.azure.net/subscriptions/00000000-0000-0000-0000-000000000000/resourcegroups/myresourcegroup/providers/Microsoft.ManagedIdentity/userAssignedIdentities/myidentity/credentials?tid=00000000-0000-0000-0000-000000000000&oid=00000000-0000-0000-0000-000000000000&aid=00000000-0000-0000-0000-000000000000",
"id": "/subscriptions/00000000-0000-0000-0000-000000000000/resourcegroups/myresourcegroup/providers/Microsoft.ManagedIdentity/userAssignedIdentities/myidentity",
"location": "eastus",
"name": "myidentity",
"principalId": "00000000-0000-0000-0000-000000000000",
"resourceGroup": "myresourcegroup",
"tags": {},
"tenantId": "00000000-0000-0000-0000-000000000000",
"type": "Microsoft.ManagedIdentity/userAssignedIdentities"
}
Save this Kubernetes manifest to a file named aadpodidentity.yaml
:
apiVersion: "aadpodidentity.k8s.io/v1"
kind: AzureIdentity
metadata:
name: <a-idname>
spec:
type: 0
ResourceID: /subscriptions/<subid>/resourcegroups/<resourcegroup>/providers/Microsoft.ManagedIdentity/userAssignedIdentities/<name>
ClientID: <clientId>
Replace the placeholders with your user identity values. Set type: 0
for user-assigned MSI or type: 1
for Service Principal.
Finally, save your changes to the file, then create the AzureIdentity
resource in your cluster:
kubectl apply -f aadpodidentity.yaml
For matching pods in the namespace, please refer to namespaced README.
Save this Kubernetes manifest to a file named aadpodidentitybinding.yaml
:
apiVersion: "aadpodidentity.k8s.io/v1"
kind: AzureIdentityBinding
metadata:
name: demo1-azure-identity-binding
spec:
AzureIdentity: <a-idname>
Selector: <label value to match>
Replace the placeholders with your values. Ensure that the AzureIdentity
name matches the one in aadpodidentity.yaml
.
Finally, save your changes to the file, then create the AzureIdentityBinding
resource in your cluster:
kubectl apply -f aadpodidentitybinding.yaml
For a pod to match an identity binding, it needs a label with the key aadpodidbinding
whose value is that of the Selector:
field in the binding. Here is an example pod with a label:
$ kubectl get po busybox0 --show-labels
NAME READY STATUS RESTARTS AGE LABELS
busybox0 1/1 Running 10 10h aadpodidbinding=select_it,app=busybox0
This pod will match the binding below:
apiVersion: "aadpodidentity.k8s.io/v1"
kind: AzureIdentityBinding
metadata:
name: test-azure-id-binding
spec:
AzureIdentity: "test-azure-identity"
Selector: "select_it"
This step is only required for user-assigned MSI.
The MIC uses the service principal credentials stored in the cluster to access Azure resources. This service principal needs Microsoft.ManagedIdentity/userAssignedIdentities/\*/assign/action
permission on the identity to work with user-assigned MSI.
If the Azure identity is in the same resource group as your AKS cluster nodes, you can skip this section. (For AKS, a resource group was added with an MC_
prefix when you created the cluster.) Otherwise, follow these steps to assign the required permissions:
-
Find the service principal used by your cluster. Please refer to the AKS docs for details. For example, if you didn't specify a service principal to the
az aks create
command, it will have generated one in the~/.azure/aksServicePrincipal.json
file. -
Assign the required permissions with the following command:
az role assignment create --role "Managed Identity Operator" --assignee <sp id> --scope <full id of the managed identity>
The NMI pods modify the nodes' iptables to intercept calls to Azure Instance Metadata endpoint. This allows NMI to insert identities assigned to a pod before executing the request on behalf of the caller.
These iptables entries will be cleaned up when the pod-identity pods are uninstalled. However, if the pods are terminated for unexpected reasons, the iptables entries can be removed with these commands on the node:
# remove the custom chain reference
iptables -t nat -D PREROUTING -j aad-metadata
# flush the custom chain
iptables -t nat -F aad-metadata
# remove the custom chain
iptables -t nat -X aad-metadata
The demonstration program illustrates how, after setting the identity and binding, the sample app can list VMs in an Azure resource group. To deploy the demo, please ensure you have completed the Prerequisites and understood the previous sections in this document.
The demo program can be found here: cmd/demo.
Here are some excerpts from the demo.
spt, err := adal.NewServicePrincipalTokenFromMSI(msiEndpoint, resource)
import "github.com/Azure/go-autorest/autorest/azure/auth"
authorizer, err := auth.NewAuthorizerFromEnvironment()
if err != nil {
logger.Errorf("failed NewAuthorizerFromEnvironment: %+v", authorizer)
return
}
vmClient := compute.NewVirtualMachinesClient(subscriptionID)
vmClient.Authorizer = authorizer
vmlist, err := vmClient.List(context.Background(), resourceGroup)
The user-assigned identity needs the "Reader" role on the resource group to list its VMs. Provide the principalId
from the user-assigned identity to this command:
az role assignment create --role Reader --assignee <principalid> --scope /subscriptions/<subscriptionid>/resourcegroups/<resourcegroup>
Update the deploy/demo/deployment.yaml
arguments with your subscription, clientID and resource group, then create the demo deployment:
kubectl apply -f deploy/demo/deployment.yaml
AAD Pod Identity has two components: the Managed Identity Controller (MIC) and the Node Managed Identity (NMI) pod.
The Managed Identity Controller (MIC) is a Kubernetes custom resource that watches for changes to pods, identities, and bindings through the Kubernetes API server. When it detects a relevant change, the MIC adds or deletes assigned identities as needed.
Specifically, when a pod is scheduled, the MIC assigns an identity to the underlying VM during the creation phase. When the pod is deleted, it removes the assigned identity from the VM. The MIC takes similar actions when identities or bindings are created or deleted.
The authorization request to fetch a Service Principal Token from an MSI endpoint is sent to a standard Instance Metadata endpoint which is redirected to the NMI pod. The redirection is accomplished by adding rules to redirect POD CIDR traffic with metadata endpoint IP on port 80 to the NMI endpoint. The NMI server identifies the pod based on the remote address of the request and then queries Kubernetes (through MIC) for a matching Azure identity. NMI then makes an Azure Active Directory Authentication Library (ADAL) request to get the token for the client id and returns it as a response. If the request had client id as part of the query, it is validated against the admin-configured client id.
Similarly, a host can make an authorization request to fetch Service Principal Token for a resource directly from the NMI host endpoint (http://127.0.0.1:2579/host/token/). The request must include the pod namespace podns
and the pod name podname
in the request header and the resource endpoint of the resource requesting the token. The NMI server identifies the pod based on the podns
and podname
in the request header and then queries k8s (through MIC) for a matching azure identity. Then NMI makes an ADAL request to get a token for the resource in the request, returning the token
and the clientid
as a response.
Here is an example cURL command:
curl http://127.0.0.1:2579/host/token/?resource=https://vault.azure.net -H "podname: nginx-flex-kv-int" -H "podns: default"
- Dive deeper into AAD Pod Identity by following the detailed Tutorial.
- Learn more about the design of AAD Pod Identity:
- Learn how to debug this project at the Debugging wiki page.
- Join us by Contributing to AAD Pod Identity.
This project has adopted the Microsoft Open Source Code of Conduct. For more information, see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.