Skip to main content

Using OPA/Gatekeeper with RabbitMQ Messaging Topology Resources

· 4 min read

Many organizations have policies around RabbitMQ usage wich they would like to enforce. This blog post explains via example how the Open Policy Agent Gatekeeper project can be used in combination with the RabbitMQ Messaging Topology Operator to manage RabbitMQ resources on Kubernetes and enforce policies on those resources by extending the Kubernetes API.

The RabbitMQ Messaging Topology Operator

The Messaging Topology Operator allows messaging topology state within a RabbitMQ cluster to be declaratively managed by extending the Kubernetes API with Custom Resource Definitions (CRD). Such messaging topology state includes vhosts, queues, exchanges, bindings, policies, federations, shovels, users, and permissions. Each of these types of objects is represented by a Kubernetes CRD, and examples of each of these Custom Resources (CRs) can be found in the documentation.

For concreteness, let us consider a queue:

kind: Queue
name: my-queue
name: my-queue
vhost: my-vhost
type: quorum
name: my-rabbit-cluster

This quourum queue is named my-queue, on the vhost my-vhost and the RabbitMQ cluster my-rabbit-cluster.

If developers are assigned roles that allow them to create Queue resources, then they are free to create any queues with any configurations they want.

Suppose we have policies regarding queues that we would like to enforce, for example, the following list of policies:

  1. The RabbitMQ Cluster must be named my-rabbit-cluster.
  2. Queues must be declared on the vhost my-vhost.

One option would be to limit developers' Kubernetes roles to prevent them from creating queues and instead institute a manual ticketing system, where a Kubernetes admin creates Queue objects for them. However, this method of policy enforcement is manual, time consuming, and prevents developers from meeting their own needs.

Enforcing Policy with Gatekeeper

Gatekeeper extends the Kubernetes API in a different way, allowing us to create webhooks to ensure Kubernetes API objects conform with policy defined via the OPA Rego language. Once we have deployed Gatekeeper in our Kuberentes cluster, we can enforce any policies we choose. In particular, we can create a constraint which allows us to enforce the above policies.

Deploying a constraint consists of three components:

  1. Gatekeeper config, informing gatekeeper what types of resources we would like to monitor.
  2. A constraint template, which lets gatekeeper know what type of constraint we would like to enforce and includes the rego policy configuration.
  3. A specific instance of the constraint.

Turning to our example, the Gatekeeper config necessary to enforce policy on RabbitMQ Queues is

kind: Config
name: config
namespace: gatekeeper-system
- group:
version: v1beta1
kind: Queue

To enforce the policies listed above, we create the following ConstraintTemplate

kind: ConstraintTemplate
name: queuevalidator
kind: QueueValidator
type: string
type: string
- target:
rego: |
package queuevalidator
violation[{"msg":msg}] {
allowedRabbit := input.parameters.rabbit
givenRabbit :=
givenRabbit != allowedRabbit
allowedVhost := input.parameters.vhost
givenVhost :=
givenVhost != allowedVhost
msg := sprintf("Rabbit Cluster must be %v, queues must be declared on vhost %v", [allowedRabbit, allowedVhost])

From this ConstraintTemplate, Gatekeeper will create a custom resource kind QueueValidator which takes two properties, a rabbit and a vhost, both strings. This allows us to configure the allowed RabbitMQ cluster name and vhost as parameters when deploying an instance of the constraint. The rego code ensures that the rabbitmqClusterReference and vhost match the specified allowed values. More generally, the rego block must include a violation function which evaluates to true when a policy violation occurs, along with a message explaining the policy violation.

We must first deploy the ConstraintTemplate and allow Gatekeeper to create the CRD for the kind QueueValidator before we can deploy a QueueValidator instance.

kind: QueueValidator
name: queue-validator
- apiGroups: [""]
kinds: ["Queue"]
- rabbit: my-rabbit-cluster
- vhost: my-vhost

With all of this configuration in place, if we attempt to create a Queue that does not conform to policy, it will be rejected by the webhook with an error.


Gatekeeper is an operator that extends the Kubernetes API to enforce policy. Together with the RabbitMQ Messaging Topology Operator, it is possible to declaratively manage RabbitMQ objects and ensure compliance via the Kubernetes API. We have shown a simple example, but the OPA language used to configure policies is highly extensible, allowing Gatekeeper to perform advanced policy management.