HashiCorp Vault

SchemaHero can be configured to use HashiCorp Vault to retrieve database credentials at runtime.

apiVersion: databases.schemahero.io/v1alpha4
kind: Database
metadata:
  name: my-db
  namespace: namespace
spec:
  connection:
    postgres:
      uri:
        valueFrom:
          vault:
            endpoint: http://<vault-endpoint>:8200
            connectionTemplate: postgres://{{ .username }}:{{ .password }}@postgres:5432/my-db
            serviceAccount: schemahero-vault
            serviceAccountNamespace: schemahero-vault
            secret: my-db
            role: schemahero
            agentInject: false
            kubernetesAuthEndpoint: /v1/auth/kubernetes_custom/login

name	description
endpoint	The http(s) endpoint of the vault API. This must be supplied in a way that's accessible from the namespace where the Database object is deployed
seviceAccount	The Kubernetes Service Account to use to authenticate with Vault
serviceAccountNamespace	The namespace that the serviceAccount is located in
secret	The name of the Vault secret to retreive
role	The role to use with Vault
agentInject	A boolean indicating if we should use the sidecar agent injection or integrate directly with the Vault API.
connectionTemplate	The Go template to use to form the connection URI. Will use a DB default if not specified
kubernetesAuthEndpoint	The Vault Kubernetes Authentication endpoint to use if this was enabled at a different path to the default. See here for more info

Agent Injector vs Vault API

SchemaHero supports integrating with Vault using the Agent Sidecar Injector or a direct integration with the Vault API.

Vault API

In most environments it's preferable to use the vault api and disable the agentInject attribute in the configuration. When using templates in Vault to build a connection string, the SchemaHero integration with the Vault API will read the template from the Vault database, and use it to create the connection string, unless a Go template is passed via the connectionTemplate parameter. The Agent Injector does not support injecting the connection string, and must be configured separately. When using the Vault API, a new secret is requested from Vault for each query (plan, apply).

Agent Injector

When the agentInject attribute is enabled, the endpoint, serviceAccount, and serviceAccountNamespace parameters are optional. In this mode, SchemaHero will simply add annotation to the Database controller to allow the Vault Sidecar Injector to add the secret via a mutating webhook admission controller.

When the agentInject attribute is enabled, the endpoint, serviceAccount, connectionTemplate and serviceAccountNamespace parameters are optional, kubernetesAuthEndpoint is ignored. In this mode, SchemaHero will simply add annotation to the Database controller to allow the Vault Sidecar Injector to add the secret via a mutating webhook admission controller. If connectionTemplate is specified if will form the Go template that's wrapped in a Consul template in the agent-inject-template annotation. As an example, having your Database object configured with:

spec:
  connection:
    postgres:
      uri:
        valueFrom:
          vault:
            connectionTemplate: postgres://{{ .username }}:{{ .password }}@postgres:5432/my-db
            secret: my-db

you would end up with a Pod annotated with:

        vault.hashicorp.com/agent-inject-template-my-db: |
          {{- with secret "database/creds/my-db" -}}
          postgres://{{ .username }}:{{ .password }}@postgres:5432/my-db
          {{- end }}

where my-db is the configured secret and postgres://{{ .username }}:{{ .password }}@postgres:5432/my-db is the configured connectionTemplate.

When using the Agent Injector option, the same secret is used for the lifetime of the controller.