Chapter 14. Using bound service account tokens
You can use bound service account tokens, which improves the ability to integrate with cloud provider identity access management (IAM) services, such as AWS IAM.
14.1. About bound service account tokens
You can use bound service account tokens to limit the scope of permissions for a given service account token. These tokens are audience and time-bound. This facilitates the authentication of a service account to an IAM role and the generation of temporary credentials mounted to a pod. You can request bound service account tokens by using volume projection and the TokenRequest API.
14.2. Configuring bound service account tokens using volume projection
You can configure pods to request bound service account tokens by using volume projection.
Prerequisites
-
You have access to the cluster as a user with the
cluster-admin
role. -
You have created a service account. This procedure assumes that the service account is named
build-robot
.
Procedure
Optional: Set the service account issuer.
This step is typically not required if the bound tokens are used only within the cluster.
WarningIf you update the
serviceAccountIssuer
field and there are bound tokens already in use, all bound tokens with the previous issuer value will be invalidated. Unless the holder of a bound token has explicit support for a change in issuer, the holder will not request a new bound token until pods have been restarted.You can force all holders to request a new bound token either by manually restarting all pods in the cluster or by performing a rolling node restart. Before performing either action, wait for a new revision of the Kubernetes API server pods to roll out with your service account issuer changes.
Edit the
cluster
Authentication
object:$ oc edit authentications cluster
Set the
spec.serviceAccountIssuer
field to the desired service account issuer value:spec: serviceAccountIssuer: https://test.default.svc 1
- 1
- This value should be a URL from which the recipient of a bound token can source the public keys necessary to verify the signature of the token. The default is
https://kubernetes.default.svc
.
- Save the file to apply the changes.
Wait for a new revision of the Kubernetes API server pods to roll out. It can take several minutes for all nodes to update to the new revision. Run the following command:
$ oc get kubeapiserver -o=jsonpath='{range .items[0].status.conditions[?(@.type=="NodeInstallerProgressing")]}{.reason}{"\n"}{.message}{"\n"}'
Review the
NodeInstallerProgressing
status condition for the Kubernetes API server to verify that all nodes are at the latest revision. The output showsAllNodesAtLatestRevision
upon successful update:AllNodesAtLatestRevision 3 nodes are at revision 12 1
- 1
- In this example, the latest revision number is
12
.
If the output shows a message similar to one of the following messages, the update is still in progress. Wait a few minutes and try again.
-
3 nodes are at revision 11; 0 nodes have achieved new revision 12
-
2 nodes are at revision 11; 1 nodes are at revision 12
Optional: Force the holder to request a new bound token either by performing a rolling node restart or by manually restarting all pods in the cluster.
Perform a rolling node restart:
WarningIt is not recommended to perform a rolling node restart if you have custom workloads running on your cluster, because it can cause a service interruption. Instead, manually restart all pods in the cluster.
Restart nodes sequentially. Wait for the node to become fully available before restarting the next node. See Rebooting a node gracefully for instructions on how to drain, restart, and mark a node as schedulable again.
Manually restart all pods in the cluster:
WarningBe aware that running this command causes a service interruption, because it deletes every running pod in every namespace. These pods will automatically restart after they are deleted.
Run the following command:
$ for I in $(oc get ns -o jsonpath='{range .items[*]} {.metadata.name}{"\n"} {end}'); \ do oc delete pods --all -n $I; \ sleep 1; \ done
Configure a pod to use a bound service account token by using volume projection.
Create a file called
pod-projected-svc-token.yaml
with the following contents:apiVersion: v1 kind: Pod metadata: name: nginx spec: containers: - image: nginx name: nginx volumeMounts: - mountPath: /var/run/secrets/tokens name: vault-token serviceAccountName: build-robot 1 volumes: - name: vault-token projected: sources: - serviceAccountToken: path: vault-token 2 expirationSeconds: 7200 3 audience: vault 4
- 1
- A reference to an existing service account.
- 2
- The path relative to the mount point of the file to project the token into.
- 3
- Optionally set the expiration of the service account token, in seconds. The default is 3600 seconds (1 hour) and must be at least 600 seconds (10 minutes). The kubelet will start trying to rotate the token if the token is older than 80 percent of its time to live or if the token is older than 24 hours.
- 4
- Optionally set the intended audience of the token. The recipient of a token should verify that the recipient identity matches the audience claim of the token, and should otherwise reject the token. The audience defaults to the identifier of the API server.
Create the pod:
$ oc create -f pod-projected-svc-token.yaml
The kubelet requests and stores the token on behalf of the pod, makes the token available to the pod at a configurable file path, and refreshes the token as it approaches expiration.
The application that uses the bound token must handle reloading the token when it rotates.
The kubelet rotates the token if it is older than 80 percent of its time to live, or if the token is older than 24 hours.
Additional resources