Skip to main content

Upgrading the RabbitMQ Kubernetes Operators

This topic describes how to upgrade the RabbitMQ Kubernetes Operators, and their components.


Upgrading the RabbitMQ Cluster Kubernetes Operator or Messaging Topology Operator involves the following effects:

  1. Updating to the new operator manifest
  2. (For specified releases) Rolling restart of RabbitmqCluster Pods

Updating to the new operator manifest

To upgrade to the new version of the operator, simply apply the new operator manifest for the desired version. For example, for the RabbitMQ Cluster Operator:

kubectl apply -f ""
# namespace/rabbitmq-system unchanged
# configured
# serviceaccount/rabbitmq-cluster-operator unchanged
# unchanged
# unchanged
# unchanged
# unchanged
# deployment.apps/rabbitmq-cluster-operator configured

This will cause the Custom Resource Definitions provided by the operator to be updated, and the Operator Pod to use the updated version of the operator container image.

New versions of the Cluster Operator typically include changes to the default version of RabbitMQ used in new clusters, if one is not specified in the spec of the RabbitmqCluster manifest. As a result, after upgrade, you will likely see that newly-created RabbitmqClusters will be running on a newer version of RabbitMQ if they do not specify a version in their spec. Note that this does not have an effect on existing RabbitmqClusters, whose manifests will have already been populated with the previous operator version's default.

Rolling restart of RabbitmqCluster Pods

In some cases, upgrading the version of the RabbitMQ Cluster Operator (not the Messaging Topology Operator) will require the Pods of RabbitmqClusters to be restarted. This is usually due to a change in the underlying PodSpec of the RabbitmqCluster that the operator creates. In this case, upgrading the Cluster Operator will lead to the RabbitmqClusters on a Kubernetes distribution undergoing a rolling restart.

If a Cluster Operator version will trigger a rolling restart of RabbitmqCluster Pods, it will be noted in the release notes for that version. For example, see