The Vault Secrets Operator creates Kubernetes secrets from Vault. The idea behind the Vault Secrets Operator is to manage secrets in Kubernetes cluster using a secure GitOps based workflow. For more information about a secure GitOps based workflow I recommend the article "Managing Secrets in Kubernetes" from Weaveworks. With the help of the Vault Secrets Operator you can commit your secrets to your git repository using a custom resource. If you apply these secrets to your Kubernetes cluster the Operator will lookup the real secret in Vault and creates the corresponding Kubernetes secret. If you are using something like Sealed Secrets for this workflow the Vault Secrets Operator can be used as replacement for this.
The Vault Secrets Operator can be installed via Helm. A list of all configurable values can be found here. The chart assumes a vault server running at http://vault:8200
, but can be overidden by specifying --set vault.address=https://vault.example.com
helm repo add ricoberger https://ricoberger.github.io/helm-charts
helm repo update
helm upgrade --install vault-secrets-operator ricoberger/vault-secrets-operator
The Vault Secrets Operator supports the KV Secrets Engine - Version 1 and KV Secrets Engine - Version 2. To create a new secret engine under a path named kvv1
and kvv2
, you can run the following command:
vault secrets enable -path=kvv1 -version=1 kv
vault secrets enable -path=kvv2 -version=2 kv
After you have enabled the secret engine, create a new policy for the Vault Secrets Operator. The operator only needs read access to the paths you want to use for your secrets. To create a new policy with the name vault-secrets-operator
and read access to the kvv1
and kvv2
path, you can run the following command:
cat <<EOF | vault policy write vault-secrets-operator -
path "kvv1/*" {
capabilities = ["read"]
}
path "kvv2/data/*" {
capabilities = ["read"]
}
EOF
To access Vault, the operator can choose between:
- Token Auth Method
- Kubernetes Auth Method
- AppRole Auth Method
- Username & Password Auth Method
- AWS Auth Method
- Azure Auth Method
- GCP Auth Method
In the next sections you can find the instructions to setup Vault for the authentication methods.
To use Token auth method for the authentication against the Vault API, you need to create a token. A token with the previously created policy can be created as follows:
vault token create -period=24h -policy=vault-secrets-operator
To use the created token you need to pass the token as an environment variable to the operator. For security reasons the operator only supports the passing of environment variables via a Kubernetes secret. The secret with the keys VAULT_TOKEN
and VAULT_TOKEN_LEASE_DURATION
(as well as optional keys VAULT_TOKEN_RENEWAL_INTERVAL
and VAULT_TOKEN_RENEWAL_RETRY_INTERVAL
to control timings for token renewals, if required. VAULT_RENEW_TOKEN
can also be set to "false"
to disable the operators token renew loop) can be created with the following command:
export VAULT_TOKEN=
export VAULT_TOKEN_LEASE_DURATION=86400
cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: Secret
metadata:
name: vault-secrets-operator
type: Opaque
data:
VAULT_TOKEN: $(echo -n "$VAULT_TOKEN" | base64)
VAULT_TOKEN_LEASE_DURATION: $(echo -n "$VAULT_TOKEN_LEASE_DURATION" | base64)
EOF
This creates a secret named vault-secrets-operator
. To use this secret in the Helm chart modify the values.yaml
file as follows:
environmentVars:
- name: VAULT_TOKEN
valueFrom:
secretKeyRef:
name: vault-secrets-operator
key: VAULT_TOKEN
- name: VAULT_TOKEN_LEASE_DURATION
valueFrom:
secretKeyRef:
name: vault-secrets-operator
key: VAULT_TOKEN_LEASE_DURATION
The recommended way to authenticate is the Kubernetes auth method, which requires a service account for communication between Vault and the Vault Secrets Operator. If you installed the operator via Helm this service account is created for you. The name of the created service account is vault-secrets-operator
. Use the following commands to set the environment variables for the activation of the Kubernetes auth method:
export VAULT_SECRETS_OPERATOR_NAMESPACE=$(kubectl get sa vault-secrets-operator -o jsonpath="{.metadata.namespace}")
export VAULT_SECRET_NAME=$(kubectl get secret vault-secrets-operator -o jsonpath="{.metadata.name}")
export SA_JWT_TOKEN=$(kubectl get secret $VAULT_SECRET_NAME -o jsonpath="{.data.token}" | base64 --decode; echo)
export SA_CA_CRT=$(kubectl get secret $VAULT_SECRET_NAME -o jsonpath="{.data['ca\.crt']}" | base64 --decode; echo)
export K8S_HOST=$(kubectl config view --minify -o jsonpath='{.clusters[0].cluster.server}')
# Verify the environment variables
env | grep -E 'VAULT_SECRETS_OPERATOR_NAMESPACE|VAULT_SECRET_NAME|SA_JWT_TOKEN|SA_CA_CRT|K8S_HOST'
# To discover the service account issuer the following commands can be used:
kubectl proxy
curl --silent http://127.0.0.1:8001/api/v1/namespaces/default/serviceaccounts/default/token -H "Content-Type: application/json" -X POST -d '{"apiVersion": "authentication.k8s.io/v1", "kind": "TokenRequest"}' | jq -r '.status.token' | cut -d . -f2 | base64 -D
Enable the Kubernetes auth method at the default path (auth/kubernetes
) and finish the configuration of Vault:
vault auth enable kubernetes
# Tell Vault how to communicate with the Kubernetes cluster
vault write auth/kubernetes/config \
issuer="https://kubernetes.default.svc.cluster.local" \
token_reviewer_jwt="$SA_JWT_TOKEN" \
kubernetes_host="$K8S_HOST" \
kubernetes_ca_cert="$SA_CA_CRT"
# Create a role named, 'vault-secrets-operator' to map Kubernetes Service Account to Vault policies and default token TTL
vault write auth/kubernetes/role/vault-secrets-operator \
bound_service_account_names="vault-secrets-operator" \
bound_service_account_namespaces="$VAULT_SECRETS_OPERATOR_NAMESPACE" \
policies=vault-secrets-operator \
ttl=24h
# If you're running Vault inside Kubernetes, you can alternatively exec into any Vault pod and run this...
# In modern versions of vault, the token / host / CA cert will be fetched automatically.
# https://github.com/hashicorp/vault-plugin-auth-kubernetes/issues/121#issuecomment-1046949951
# vault write auth/kubernetes/config \
# kubernetes_host=https://${KUBERNETES_PORT_443_TCP_ADDR}:443
When you deploy the Vault Secrets Operator via Helm chart you have to set the vault.authMethod
property to kubernetes
in the values.yaml
file, to use the Kubernetes auth method instead of the default Token auth methods.
vault:
authMethod: kubernetes
To use AppRole auth method for the authentication against the Vault API, you need to create a new AppRole.
# Enable AppRole auth method:
vault auth enable approle
# AppRole with the previously created policy can be created as follows:
vault write auth/approle/role/vault-secrets-operator \
token_policies=vault-secrets-operator
# Get AppRole ID:
vault read auth/approle/role/vault-secrets-operator/role-id
# Create a new secret for AppRole:
vault write -f auth/approle/role/vault-secrets-operator/secret-id
Use the following commands to set the environment variables for the activation of the AppRole auth method:
export VAULT_AUTH_METHOD=approle
export VAULT_ROLE_ID=
export VAULT_SECRET_ID=
export VAULT_TOKEN_MAX_TTL=86400
When you deploy the Vault Secrets Operator via Helm chart you have to set the vault.authMethod
property to approle
in the values.yaml
file, to use the AppRole auth method instead of the default Token auth method.
vault:
authMethod: approle
Set VAULT_TOKEN_MAX_TTL
(default: 16 days) same or lower than the token_max_ttl
of the AppRole (Vault default: 32 days) to ensure reauthentication in time.
Mounting the vault ROLE_ID and SECRET_ID secrets as volumes is supported. It requires image.volumeMounts
to be populated, VAULT_ROLE_ID_PATH
and VAULT_SECRET_ID_PATH
to be set in environmentVars
(or export
as shell variables), and volumes
to be populated. See example below:
NOTE: image.volumeMounts[].mountPath
must match environmentVars[].value
for the respective ROLE_ID or SECRET_ID. Reference Kubernetes: Using Secrets as files from a Pod
image:
volumeMounts:
- name: vault-role-id
mountPath: "/etc/vault/role/"
readOnly: true
- name: vault-secret-id
mountPath: "/etc/vault/secret/"
readOnly: true
environmentVars:
- name: VAULT_ROLE_ID_PATH
value: "/etc/vault/role/id"
- name: VAULT_SECRET_ID_PATH
value: "/etc/vault/secret/id"
volumes:
- name: vault-role-id
secret:
secretName: vault-secrets-operator
items:
- key: VAULT_ROLE_ID
path: "id"
- name: vault-secret-id
secret:
secretName: vault-secrets-operator
items:
- key: VAULT_SECRET_ID
path: "id"
To use username and password auth method for the authentication against the Vault API, you need to create a new user.
# Enable userpass auth method
vault auth enable userpass
# Create a user if not already created
vault write auth/userpass/users/<username> password=<password> policies=<policies>
Use the following commands to set the environment variables for the activation of the UserPass auth method:
export VAULT_AUTH_METHOD=userpass
export VAULT_USER=
export VAULT_PASSWORD=
export VAULT_TOKEN_MAX_TTL=120
When you deploy the Vault Secrets Operator via Helm chart you have to set the vault.authMethod
property to userpass
in the values.yaml
file, to use the UserPass auth method instead of the default Token auth method.
vault:
authMethod: userpass
You can use either ec2 or iam auth types on eks clusters to authenticate against the Vault API. here Then you can enable the auth method with the following environment variables:
export VAULT_AUTH_METHOD=aws
export VAULT_AWS_PATH=auth/aws
export VAULT_AWS_ROLE=vault-secrets-operator
export VAULT_AWS_AUTH_TYPE=iam
If you deploy the Vault Secrets Operator via Helm you have to set the vault.authMethod
, vault.awsPath
, vault.awsRole
and vault.awsAuthType
values in the values.yaml
file.
You can use the managed system identity provided on aks cluster to authenticate against the Vault API, to do that you will need to setup an auth backend as described here Then you can setup the auth method with the following environment variables:
export VAULT_AUTH_METHOD=azure
export VAULT_AZURE_PATH=auth/azure
export VAULT_AZURE_ROLE=default
export VAULT_AZURE_ISSCALESET=true # Set this to true if the kubernetes nodes are in a vmss and not isolated vm (default in aks)
If you deploy the Vault Secrets Operator via Helm you have to set the vault.authMethod
, vault.azurepath
, vault.azureRole
, vault.azureScaleset
values in the values.yaml
file.
You can use either gce or iam auth types on gke clusters to authenticate against the Vault API. here Then you can enable the auth method with the following environment variables:
export VAULT_AUTH_METHOD=gcp
export VAULT_GCP_PATH=auth/gcp
export VAULT_GCP_ROLE=vault-secrets-operator
export VAULT_GCP_AUTH_TYPE=iam
If you deploy the Vault Secrets Operator via Helm you have to set the vault.authMethod
, vault.gcpPath
, vault.gcpRole
and vault.gcpAuthType
values in the values.yaml
file.
Create two Vault secrets example-vaultsecret
:
vault kv put kvv1/example-vaultsecret foo=bar hello=world
vault kv put kvv2/example-vaultsecret foo=bar
vault kv put kvv2/example-vaultsecret hello=world
vault kv put kvv2/example-vaultsecret foo=bar hello=world
Deploy the custom resource kvv1-example-vaultsecret
to your Kubernetes cluster:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv1-example-vaultsecret
spec:
keys:
- foo
path: kvv1/example-vaultsecret
type: Opaque
The Vault Secrets Operator creates a Kubernetes secret named kvv1-example-vaultsecret
with the type Opaque
from this CR:
apiVersion: v1
data:
foo: YmFy
kind: Secret
metadata:
labels:
created-by: vault-secrets-operator
name: kvv1-example-vaultsecret
type: Opaque
You can also omit the keys
spec to create a Kubernetes secret which contains all keys from the Vault secret:
apiVersion: v1
data:
foo: YmFy
hello: d29ybGQ=
kind: Secret
metadata:
labels:
created-by: vault-secrets-operator
name: kvv1-example-vaultsecret
type: Opaque
To deploy a custom resource kvv2-example-vaultsecret
, which uses the secret from the KV Secrets Engine - Version 2 you can use the following:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv2-example-vaultsecret
spec:
path: kvv2/example-vaultsecret
type: Opaque
The Vault Secrets Operator will create a secret which looks like the following:
apiVersion: v1
data:
foo: YmFy
hello: d29ybGQ=
kind: Secret
metadata:
labels:
created-by: vault-secrets-operator
name: kvv2-example-vaultsecret
type: Opaque
For secrets using the KVv2 secret engine you can also specify the version of the secret you want to deploy:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv2-example-vaultsecret
spec:
path: kvv2/example-vaultsecret
type: Opaque
version: 2
The resulting Kubernetes secret will be:
apiVersion: v1
data:
hello: d29ybGQ=
kind: Secret
metadata:
labels:
created-by: vault-secrets-operator
name: kvv2-example-vaultsecret
type: Opaque
The spec.type
and spec.keys
fields are handled in the same way for both versions of the KV secret engine. The spec.version
field is only processed, when the secret is saved under a KVv2 secret engine. If you specified the VAULT_RECONCILIATION_TIME
environment variable with a value greater than 0
every secret is reconciled after the given time (in seconds). This means, when you do not specify spec.version
, the Kubernetes secret will be automatically updated if the Vault secret changes. To set the VAULT_RECONCILIATION_TIME
environment variable in the Helm chart the vault.reconciliationTime
value can be used.
The binary data stored in vault requires base64 encoding. the
spec.isBinary
can be used to prevent such data get base64 encoded again when store as secret in k8s.
For example, let's set foo
to the bar
in base64 encoded format (i.e. YmFyCg==).
vault kv put kvv1/example-vaultsecret foo=YmFyCg==
You can specify spec.isBinary
to indicate this is a binary data which is already in base64 encoded format:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv1-example-vaultsecret
spec:
keys:
- foo
isBinary: true
path: kvv1/example-vaultsecret
type: Opaque
The resulting Kubernetes secret will be:
apiVersion: v1
data:
foo: YmFyCg==
kind: Secret
metadata:
labels:
created-by: vault-secrets-operator
name: kvv1-example-vaultsecret
type: Opaque
The value for foo
stays as YmFyCg==
which does not get base64 encoded again.
It is also possible to change the default reconciliation strategy from Replace
to Merge
via the reconcileStrategy
key in the CRD. For the default Replace
strategy the complete secret is replaced. If you have an existing secret you can choose the Merge
strategy to add the keys from Vault to the existing secret.
When straight-forward secrets are not sufficient, and the target secrets need to be formatted in a certain way, you can use basic templating to format the secrets. There are multiple uses for this:
- Generate URIs which contain secrets
- Format secrets in a specific way, for example when using the Helm Operator which can use secrets as a source for its Helm chart parameterisation, but they have to be in YAML format wrapped inside a secret, like
secretGenerator
from Kustomize also generates.
To do this, specify keys under spec.templates
, containing a valid template string.
When templates
is defined, the standard generation of secrets is disabled, and only the defined templates will be generated.
The templating uses the standard Go templating engine, also used in tools such as Helm or Gomplate. The main differentiator here is that the {%
and %}
delimiters are used to prevent conflicts with standard Go templating tools such as Helm, which use {{
and }}
for this.
The available functions during templating are the set offered by the Sprig library (similar to Helm, but different from Gomplate), excluding the following functions for security-reasons or their non-idempotent nature to avoid reconciliation problems:
genPrivateKey
genCA
genSelfSignedCert
genSignedCert
htpasswd
getHostByName
- Random functions
- Date/time functionality
- Environment variable functions (for security reasons)
The context available in the templating engine contains the following items:
.Secrets
: Map with all the secrets fetched from vault. Key = secret name, Value = secret value.Vault
: Contains misc info about the Vault setup.Vault.Address
: configured address of the Vault instance.Vault.Path
: path of the Vault secret that was fetched
.Namespace
: Namespace where the custom resource instance was deployed..Labels
: access to the labels of the custom resource instance.Annotations
: access to the annotations of the custom resource instance
An example of a URI formatting secret:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv1-example-vaultsecret
annotations:
redisdb: "0"
spec:
keys:
- foo
- bar
path: kvv1/example-vaultsecret
templates:
fooUri: "https://user:{% .Secrets.foo %}@{% .Namespace %}.somesite.tld/api"
barUri: "redis://{% .Secrets.bar %}@redis/{% .Annotations.redisdb %}"
type: Opaque
The resulting secret will look like:
apiVersion: v1
data:
fooUri: aHR0cHM6Ly91c2VyOmZvb0BuYW1lc3BhY2UuLnNvbWVzaXRlLnRsZC9hcGkK
barUri: cmVkaXM6Ly9iYXJAcmVkaXMvMAo=
kind: Secret
metadata:
labels:
created-by: vault-secrets-operator
name: kvv1-example-vaultsecret
type: Opaque
This is a more advanced example for a secret that can be used by HelmOperator as valuesFrom[].secretKeyRef
:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv1-example-vaultsecret
spec:
keys:
- foo
- bar
- baz
path: kvv1/example-vaultsecret
templates:
values.yaml: |-
secrets:
{%- range $k, $v := .Secrets %}
{% $k %}: {% $v | quote -%}
{% end %}
type: Opaque
This will loop over all secrets fetched from Vault, and set the vault.yaml
key to a string like this:
secrets:
foo: "foovalue"
bar: "barvalue"
baz: "bazvalue
- All secrets data is converted to string before being passed to the templating engine, so using binary data will not work well, or at least be unpredictable.
You can generate certificates using the PKI Engine like so:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: test-pki
spec:
path: pki
secretEngine: pki
role: example-dot-com
engineOptions:
common_name: www.my-website.com
ttl: "5d"
type: Opaque
You can pass any of the parameters supported by the PKI engine in engineOptions
.
A list is available here: https://www.vaultproject.io/api-docs/secret/pki#parameters-15
This will generate the following secret:
apiVersion: v1
kind: Secret
metadata:
name: test-pki
data:
certificate: LS0tLS1CRUdJTiBDRVJUSUZJQ0FURS0tLS0tLi4u
expiration: MTY0OTc2OTIwMg==
issuing_ca: LS0tLS1CRUdJTiBDRVJUSUZJQ0FURS0tLS0tLi4u
private_key: LS0tLS1CRUdJTiBSU0EgUFJJVkFURSBLRVktLS0tLS4uLg==
private_key_type: cnNh
serial_number: MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA6MDA=
type: Opaque
Like with the secret engine, you can template the resulting Kubernetes secret:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: test-pki
spec:
path: pki
secretEngine: pki
role: example-dot-com
engineOptions:
common_name: www.my-website.com
templates:
tls.crt: "{% .Secrets.certificate %}"
tls.key: "{% .Secrets.private_key %}"
ca.crt: "{% .Secrets.issuing_ca %}"
type: Opaque
The following fields are available:
certificate
expiration
issuing_ca
private_key
private_key_type
serial_number
Certificate are renewed before expiration. You can set how long before expiration you want to renew by setting the
VAULT_PKI_RENEW
environment variable. The default is 1 hour.
It is possible to not set the VAULT_KUBERNETES_ROLE
(vault.kubernetesRole
value in the Helm chart) and instead specify the Vault Role at the CR. This allows you to to use different Vault Roles within one Vault Secrets Operator instance.
The Vault Role is set via the vaultRole
property in the VaultSecret CR:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv1-example-vaultsecret
spec:
vaultRole: my-custom-vault-role
path: kvv1/example-vaultsecret
type: Opaque
Note: This option is only available for the kubernetes auth method and all roles must be added to the auth method before they are used by the operator.
Vault Namespaces is a set of features within Vault Enterprise that allows Vault environments to support Secure Multi-tenancy (or SMT) within a single Vault infrastructure.
The Vault Namespace, which should be used for the authentication of the operator against Vault can be specified via the VAULT_NAMESPACE
environment variable. In the Helm chart this value can be provided as follows:
environmentVars:
- name: VAULT_NAMESPACE
value: "my/root/ns"
The operator also supports nested Namespaces. When the VAULT_NAMESPACE
is set, it is also possible to specify a namespace via the vaultNamespace
field in the VaultSecret CR:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: kvv1-example-vaultsecret
spec:
vaultNamespace: team1
path: kvv1/example-vaultsecret
type: Opaque
The Vault Namespace, which is used to get the secret in the above example will be my/root/ns/team1
.
The operator can also be restricted to only reconcile secrets where the spec.vaultNamespace
field is the same as the VAULT_NAMESPACE
environment variable. For this the VAULT_RESTRICT_NAMESPACE
environment variable must be set to true
. When this feature is enabled the operator can not be used with nested namespaces.
The operator will propagate all labels found on the VaultSecret
to the actual secret. So if a given label is needed on the resulting secret it can be added like in the following example:
apiVersion: ricoberger.de/v1alpha1
kind: VaultSecret
metadata:
name: example-vaultsecret
labels:
my-custom-label: my-custom-label-value
spec:
path: path/to/example-vaultsecret
type: Opaque
This would result in the following secret:
apiVersion: v1
data:
...
kind: Secret
metadata:
labels:
created-by: vault-secrets-operator
my-custom-label: my-custom-label-value
name: example-vaultsecret
type: Opaque
After modifying the *_types.go
file always run the following command to update the generated code for that resource type:
make generate
The above Makefile target will invoke the controller-gen utility to update the
api/v1alpha1/zz_generated.deepcopy.go
file to ensure our API's Go type definitons implement the runtime.Object
interface that all Kind types must implement.
Once the API is defined with spec/status fields and CRD validation markers, the CRD manifests can be generated and updated with the following command:
make manifests
This Makefile target will invoke controller-gen to generate the CRD manifest at charts/vault-secrets-operator/crds/ricoberger.de_vaultsecrets.yaml
.
Deploy the CRD and run the operator locally with the default Kubernetes config file present at $HOME/.kube/config
:
export VAULT_ADDRESS=
export VAULT_AUTH_METHOD=token
export VAULT_TOKEN=
export VAULT_TOKEN_LEASE_DURATION=86400
export VAULT_RECONCILIATION_TIME=180
make run