KubeVault’s built-in CSI driver has been removed in favor of
Secrets Store CSI driver for Kubernetes secrets.
Mount GCP Secrets using CSI Driver
At first, you need to have a Kubernetes 1.14 or later cluster, and the kubectl command-line tool must be configured to communicate with your cluster. If you do not already have a cluster, you can create one by using kind. To check the version of your cluster, run:
$ kubectl version --short
Client Version: v1.16.2
Server Version: v1.14.0
Before you begin:
- Install KubeVault operator in your cluster from here.
- Install Secrets Store CSI driver for Kubernetes secrets in your cluster from here.
To keep things isolated, we are going to use a separate namespace called demo throughout this tutorial.
$ kubectl create ns demo
namespace/demo created
Note: YAML files used in this tutorial stored in examples folder in GitHub repository KubeVault/docs
Vault Server
If you don’t have a Vault Server, you can deploy it by using the KubeVault operator.
The KubeVault operator can manage policies and secret engines of Vault servers which are not provisioned by the KubeVault operator. You need to configure both the Vault server and the cluster so that the KubeVault operator can communicate with your Vault server.
Now, we have the AppBinding that contains connection and authentication information about the Vault server. And we also have the service account that the Vault server can authenticate.
$ kubectl get serviceaccounts -n demo
NAME SECRETS AGE
vault 1 20h
$ kubectl get appbinding -n demo
NAME AGE
vault 50m
$ kubectl get appbinding -n demo vault -o yaml
apiVersion: appcatalog.appscode.com/v1alpha1
kind: AppBinding
metadata:
name: vault
namespace: demo
spec:
clientConfig:
caBundle: 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
service:
name: vault
port: 8200
scheme: HTTPS
parameters:
apiVersion: config.kubevault.com/v1alpha1
kind: VaultServerConfiguration
path: kubernetes
vaultRole: vault-policy-controller
kubernetes:
serviceAccountName: vault
tokenReviewerServiceAccountName: vault-k8s-token-reviewer
usePodServiceAccountForCSIDriver: true
The following steps are required to enable and configure the GCP secrets engine in the Vault server.
There are two ways to configure the Vault server. You can either use the KubeVault operator or the Vault CLI to manually configure a Vault server.
Using KubeVault operator
You need to be familiar with the following CRDs:
Let’s enable and configure GCP secret engine by deploying the following SecretEngine yaml:
apiVersion: engine.kubevault.com/v1alpha1
kind: SecretEngine
metadata:
name: gcp-engine
namespace: demo
spec:
vaultRef:
name: vault
gcp:
credentialSecret: gcp-cred
To configure the GCP secret engine, you need to provide google service account credentials through a Kubernetes secret.
apiVersion: v1
kind: Secret
metadata:
name: gcp-cred
namespace: demo
data:
sa.json: eyJtc2ciOiJleGFtcGxlIn0= ## base64 encoded google service account credential
Let’s deploy SecretEngine:
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/gcpCred.yaml
secret/gcp-cred created
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/gcpSecretEngine.yaml
secretengine.engine.kubevault.com/gcp-engine created
Wait till the status become Success:
$ kubectl get secretengines -n demo
NAME STATUS
gcp-engine Success
Create GCP roleset using the following GCPRole yaml:
apiVersion: engine.kubevault.com/v1alpha1
kind: GCPRole
metadata:
name: gcp-role
namespace: demo
spec:
vaultRef:
name: vault
secretType: access_token
project: ackube
bindings: |
resource "//cloudresourcemanager.googleapis.com/projects/ackube" {
roles = ["roles/viewer"]
}
tokenScopes:
- https://www.googleapis.com/auth/cloud-platform
Let’s deploy GCPRole:
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/gcpRole.yaml
gcprole.engine.kubevault.com/gcp-role created
$ kubectl get gcprole -n demo
NAME STATUS
gcp-role Success
You can also check from Vault that the roleset is created.
To resolve the naming conflict, name of the roleset in Vault will follow this format: k8s.{clusterName}.{metadata.namespace}.{metadata.name}.
Don’t have Vault CLI? Download and configure it as described here
$ vault list gcp/roleset
Keys
----
k8s.-.demo.gcp-role
$ vault read gcp/roleset/k8s.-.demo.gcp-role
Key Value
--- -----
bindings map[//cloudresourcemanager.googleapis.com/projects/ackube:[roles/viewer]]
project ackube
secret_type access_token
service_account_email vaultk8s---demo-gcp-424523423@ackube.iam.gserviceaccount.com
token_scopes [https://www.googleapis.com/auth/cloud-platform]
Using Vault CLI
You can also use Vault CLI to enable and configure the GCP secret engine.
Don’t have Vault CLI? Download and configure it as described here
To generate secret from the GCP secret engine, you have to perform the following steps.
- Enable GCP Secret Engine: To enable the GCP secret engine, run the following command.
$ vault secrets enable gcp
Success! Enabled the gcp secrets engine at: gcp/
- Configure the secrets engine: Configure the secrets engine with google service account credentials
$ vault write gcp/config credentials=@/home/user/Downloads/ackube-sa.json
Success! Data written to: gcp/config
- Configure a roleset: Configure a roleset that generates OAuth2 access tokens
$ vault write gcp/roleset/k8s.-.demo.gcp-role \
project="ackube" \
secret_type="access_token" \
token_scopes="https://www.googleapis.com/auth/cloud-platform" \
bindings='resource "//cloudresourcemanager.googleapis.com/projects/ackube" {
roles = ["roles/viewer"]
}'
Success! Data written to: gcp/roleset/k8s.-.demo.gcp-role
$ vault read gcp/roleset/k8s.-.demo.gcp-role
Key Value
--- -----
bindings map[//cloudresourcemanager.googleapis.com/projects/ackube:[roles/viewer]]
project ackube
secret_type access_token
service_account_email vaultk8s---demo-gcp-424523423@ackube.iam.gserviceaccount.com
token_scopes [https://www.googleapis.com/auth/cloud-platform]
If you use Vault CLI to enable and configure the GCP secret engine then you need to update the vault policy for the service account ‘vault’ created during vault server configuration and add the permission to read at “gcp/roleset/*” with previous permissions. That is why it is recommended to use the KubeVault operator because the operator updates the policies automatically when needed.
Find how to update the policy for service account in here.
Mount secrets into a Kubernetes pod
Since Kubernetes 1.14, storage.k8s.io/v1beta1 CSINode and CSIDriver objects were introduced. Let’s check CSIDriver and CSINode are available or not.
$ kubectl get csidrivers
NAME CREATED AT
secrets.csi.kubevault.com 2019-12-09T04:32:50Z
$ kubectl get csinodes
NAME CREATED AT
2gb-pool-57jj7 2019-12-09T04:32:52Z
2gb-pool-jrvtj 2019-12-09T04:32:58Z
So, we can create StorageClass now.
Create StorageClass
Create StorageClass object with the following content:
kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
name: vault-gcp-storage
annotations:
storageclass.kubernetes.io/is-default-class: "false"
provisioner: secrets.csi.kubevault.com
parameters:
ref: demo/vault # namespace/AppBinding, we created this in previous step
engine: GCP # vault engine name
role: k8s.-.demo.gcp-role # roleset name created during vault configuration
path: gcp # specifies the secret engine path, default is gcp
secret_type: access_token # Specifies the type of secret generated for this role set, i.e. access_token or service_account_key
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/storageClass.yaml
storageclass.storage.k8s.io/vault-gcp-storage created
Note: you can also provide key_algorithm and key_type fields as parameters when secret_type is service_account_key.
Test & Verify
Let’s create a separate namespace called trial for testing purpose.
$ kubectl create ns trial
namespace/trail created
Create PVC
Create a PersistentVolumeClaim with the following data. This makes sure a volume will be created and provisioned on your behalf.
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
name: csi-pvc-gcp
namespace: trial
spec:
accessModes:
- ReadWriteOnce
resources:
requests:
storage: 100Mi
storageClassName: vault-gcp-storage
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/pvc.yaml
persistentvolumeclaim/csi-pvc-gcp created
Create VaultPolicy and VaultPolicyBinding for Pod’s Service Account
Let’s say pod’s service account name is pod-sa located in trial namespace. We need to create a VaultPolicy and a VaultPolicyBinding so that the pod has access to read secrets from the Vault server.
apiVersion: policy.kubevault.com/v1alpha1
kind: VaultPolicy
metadata:
name: gcp-se-policy
namespace: demo
spec:
vaultRef:
name: vault
# Here, gcp secret engine is enabled at "gcp".
# If the path was "demo-se", policy should be like
# path "demo-se/*" {}.
policyDocument: |
path "gcp/*" {
capabilities = ["create", "read"]
}
---
apiVersion: policy.kubevault.com/v1alpha1
kind: VaultPolicyBinding
metadata:
name: gcp-se-role
namespace: demo
spec:
vaultRef:
name: vault
policies:
- ref: gcp-se-policy
subjectRef:
kubernetes:
serviceAccountNames:
- "pod-sa"
serviceAccountNamespaces:
- "trial"
Let’s create VaultPolicy and VaultPolicyBinding:
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/vaultPolicy.yaml
vaultpolicy.policy.kubevault.com/gcp-se-policy created
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/vaultPolicyBinding.yaml
vaultpolicybinding.policy.kubevault.com/gcp-se-role created
Check if the VaultPolicy and the VaultPolicyBinding are successfully registered to the Vault server:
$ kubectl get vaultpolicy -n demo
NAME STATUS AGE
gcp-se-policy Success 8s
$ kubectl get vaultpolicybinding -n demo
NAME STATUS AGE
gcp-se-role Success 10s
Create Service Account for Pod
Let’s create the service account pod-sa which was used in VaultPolicyBinding. When a VaultPolicyBinding object is created, the KubeVault operator create an auth role in the Vault server. The role name is generated by the following naming format: k8s.(clusterName or -).namespace.name. Here, it is k8s.-.demo.gcp-se-role. We need to provide the auth role name as service account annotations while creating the service account. If the annotation secrets.csi.kubevault.com/vault-role is not provided, the CSI driver will not be able to perform authentication to the Vault.
apiVersion: v1
kind: ServiceAccount
metadata:
name: pod-sa
namespace: trial
annotations:
secrets.csi.kubevault.com/vault-role: k8s.-.demo.gcp-se-role
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/podServiceAccount.yaml
serviceaccount/pod-sa created
Create Pod
Now we can create a Pod which refers to this volume. When the Pod is created, the volume will be attached, formatted and mounted to the specific container.
apiVersion: v1
kind: Pod
metadata:
name: mypod
namespace: trial
spec:
containers:
- name: mypod
image: busybox
command:
- sleep
- "3600"
volumeMounts:
- name: my-vault-volume
mountPath: "/etc/gcp"
readOnly: true
serviceAccountName: pod-sa # service account that was created
- name: my-vault-volume
persistentVolumeClaim:
claimName: csi-pvc-gcp
$ kubectl apply -f docs/examples/guides/secret-engines/gcp/pod.yaml
pod/mypod created
Check if the Pod is running successfully, by running:
$ kubectl get pods -n trial
NAME READY STATUS RESTARTS AGE
mypod 1/1 Running 0 11s
Verify Secret
If the Pod is running successfully, then check inside the app container by running
$ kubectl exec -it -n trial mypod sh
/ # ls /etc/gcp
expires_at_seconds token token_ttl
/ # cat /etc/gcp/token
ya29.c.Kl20BwwWtb6DoTjY4-eSVgQQq.......
So, we can see that the secret token is mounted into the pod, where the secret key is mounted as file and the value is the content of that file.
Cleaning up
To clean up the Kubernetes resources created by this tutorial, run:
$ kubectl delete ns demo
namespace "demo" deleted
$ kubectl delete ns trial
namespace "trial" deleted
