Skip to content

Kubernetes Operator to sync secrets between different secret backends and Kubernetes

License

Notifications You must be signed in to change notification settings

digitalis-io/vals-operator

Repository files navigation

vals-operator

Vals-Operator

CI GoDoc Go Report Card GitHub GitHub release (latest by date) Artifact Hub

Vals-Operator is a Kubernetes operator that integrates external

Here at Digitalis we love vals, it's a tool we use daily to keep secrets stored securely. Inspired by it, we have created an operator to manage Kubernetes secrets. As Digitalis and our sister company AxonOps are data companies, we also added a set of features tailored for running databases.

vals-operator syncs secrets from any secrets store supported by vals into Kubernetes. Also, vals-operator supports database secrets as provider by the HashiCorp Vault Secret Engine.

Demo

You can watch this brief video on how it works:

YouTube

Mirroring secrets

We have also added the ability to copy secrets between namespaces. It uses the format ref+k8s://namespace/secret#key. This way you can keep secrets generated in one namespace in sync with any other namespace in the cluster.

Installation

You can use the helm chart to install vals-operator. First of all, add the repository to your helm installation:

helm repo add digitalis https://digitalis-io.github.io/helm-charts

You will need to provide the configuration to access the secrets store you decided on via either environment variables pre existing secrets.

# Example for Vault
helm upgrade --install vals-operator --create-namespace -n vals-operator \
  --set "env[0].name=VAULT_ROLE_ID,env[0].value=vals-operator" \
  --set "env[1].name=VAULT_SECRET_ID,env[1].value=my-secret-id" \
  --set "env[2].name=VAULT_ADDR,env[2].value=https://vault:8200"
  digitalis/vals-operator

# Example for AWS using a secret
kubectl create secret generic -n vals-operator aws-creds \
  --from-literal=AWS_ACCESS_KEY_ID=foo \
  --from-literal=AWS_SECRET_ACCESS_KEY=bar \
  --from-literal=AWS_DEFAULT_REGION=us-west-2

helm upgrade --install vals-operator --create-namespace -n vals-operator \
  --set "secretEnv[0].secretRef.name=aws-creds"  \
  digitalis/vals-operator

# Another example using a Google Cloud service account
kubectl create secret generic -n vals-operator google-creds \
  --from-file=credentials.json=/path/to/service_account.json

helm upgrade --install vals-operator --create-namespace -n vals-operator \
  --set "env[0].name=GOOGLE_APPLICATION_CREDENTIALS,env[0].value=/secret/credentials.json" \
  --set "env[1].name=GCP_PROJECT,env[1].value=my_project" \
  --set "volumes[0].name=creds,volumes[0].secret.secretName=google-creds" \
  --set "volumeMounts[0].name=creds,volumeMounts[0].mountPath=/secret" \
  digitalis/vals-operator

ℹ️ Check out the documentation for further details and examples including EKS integration.

HashiCorp Vault Authentication

If you're using Vault as backend you can also enable the Kubernetes Auth login method. Refer to the HashiCorp documentation on creating a role.

You will need to add two additional environment variables to the vals-operator installation:

  • VAULT_ROLE_ID: required to enable Kubernetes Login
  • VAULT_ADDR: URL to the Vault server, ie, http://vault:8200
  • VAULT_LOGIN_USER and VAULT_LOGIN_PASSWORD: to use userpass authentication (insecure, not recommended)
  • VAULT_APP_ROLE and VAULT_SECRET_ID: to use approle authentication

Usage

apiVersion: digitalis.io/v1
kind: ValsSecret
metadata:
  name: vals-secret-sample
  labels:
    owner: digitalis.io
spec:
  name: my-secret # Optional, default is the resource name
  ttl: 3600       # Optional, default is 5 minutes. The secret will be checked at every "reconcile period". See below.
  type: Opaque    # Default type, others supported
  data:
    username:
      ref: ref+vault://secret/database/username
      encoding: text
    password:
      ref: ref+vault://secret/database/password
      encoding: text
    ssh:
      ref: ref+vault://secret/database/ssh-private-key
      encoding: base64
    aws-user:
      ref: ref+awssecrets://kube/test#username
    aws-pass:
      ref: ref+awssecrets://kube/test#password
    ns-secret:
      ref: ref+k8s://namespace/secret#key
    plain-text:
      ref: literal_name # this is not processed by any secrets agent but is added to the secret as a literal string
  template:
    config.yaml: |
      # Config generated by Vals-Operator on {{ now | date "2006-01-02" }}
      username: {{.username}}
      password: {{.password}}
      {{- if .url }}
      url: {{ .url | lower }}
      {{ end }}
  rollout: # optional: run a `rollout` to make the pods use new secret
    - kind: Deployment
      name: myapp

The example above will create a secret named my-secret and get the values from the different sources. The secret will be kept in sync against the backed secrets store.

The TTL is optional and used to decrease the number of times the operator calls the backend secrets store as some of them such as AWS Secrets Manager will incur a cost.

The default encoding is text but you can change it to base64 per secret reference. This way you can, for example, base64 encode large configuration files. If you omit the ref+ prefix vals-operator will not process the string and it will be added to the secret as as literal string.

You may also use GoLang templates to format a secret. You can inject as variables any of the keys referenced in the data section to format, for example, a configuration file. The sprig functions are supported.

Vault database credentials


NOTE: Vault >= 1.10 is required for this feature to work


A great feature in HashiCorp Vault is the generate database credentials dynamically. The missing part is you need these credentials in Kubernertes where your applications are. This is why we have added a new resource definition to do just that:

apiVersion: digitalis.io/v1beta1
kind: DbSecret
metadata:
  name: cassandra
spec:
  renew: true # this is the default, otherwise a new credential will be generated every time
  vault:
    role: readonly
    mount: cass000
  template: # optional: change the secret format
    CASSANDRA_USERNAME: "{{ .username }}"
    CASSANDRA_PASSWORD: "{{ .password }}"
  rollout: # optional: run a `rollout` to make the pods use new credentials
    - kind: Deployment
      name: cassandra-client
    - kind: StatefulSet
      name: cassandra-client-other

Advance config: password rotation

If you're running a database you may want to keep the secrets in sync between your secrets store, Kubernetes and the database. This can be handy for password rotation to ensure the clients don't use the same password all the time. Please be aware your client must suppport re-reading the secret and reconnecting whenever it is updated.

We don't yet support TLS, we'll add it to future releases.

---
apiVersion: digitalis.io/v1
kind: ValsSecret
metadata:
  name: vals-secret-sample
  labels:
    owner: digitalis.io
spec:
  name: my-secret # Optional, default is the resource name
  ttl: 10         # Optional, default is 0. The secret will be checked at every "reconcile period". See below.
  type: Opaque    # Default type, others supported
  data:
    username:
      ref: ref+gcpsecrets://databases/test#username
      encoding: text
    password:
      ref: ref+gcpsecrets://databases/test#password
      encoding: text
  databases:
    - driver: cassandra
      loginCredentials:
        secretName: cassandra-creds # secret containing the username and password to access the DB and run the below query
        usernameKey: username       # in the secret, which key contains the username (default `cassandra`)
        passwordKey: password       # in the secret, which key contains the password
      port: 9042
      usernameKey: username
      passwordKey: password
      hosts:                        # list all your cassandra nodes here
        - cassandra01
        - cassandra02
    - driver: postgres
      loginCredentials:
        secretName: postgres-creds
        usernameKey: username
        passwordKey: password
      port: 5432
      usernameKey: username
      passwordKey: password
      hosts:
        - postgres
    - driver: mysql
      loginCredentials:
        secretName: mysql-creds
        namespace: mysql-server
        passwordKey: mysql-root-password # if username is omitted it defaults to `mysql`
      port: 3306
      usernameKey: username
      passwordKey: password
      userHost: "%"                     # default
      hosts:
        - mysql
    - driver: elastic
      loginCredentials:
        secretName: elastic-creds
        namespace: elastic-server
        usernameKey: username           # the username defaults to 'elastic' if not provided
        passwordKey: password
      port: 9200
      usernameKey: username
      passwordKey: password
      hosts:
        - my-elastic                    # this would be converted to http://my-elastic:9200
        - https://my-other-elastic:9200 # provide full URL instead