Skip to content
Branch: master
Find file History
michaelklishin Introduce a supervision tree for node monitoring
Otherwise the monitor process would not be restarted
should it fail.
Latest commit 446ef09 Dec 9, 2019

Deploy RabbitMQ on Kubernetes with the Kubernetes Peer Discovery Plugin to Minikube

This is an example that demonstrates a RabbitMQ deployment on Kubernetes with peer discovery via rabbitmq-peer-discovery-k8s plugin.

Production (Non-)Suitability

Some values in this example may or may not be optimal for your deployment. We encourage users to get familiar with the RabbitMQ Peer Discovery guide, RabbitMQ Production Checklist and the rest of RabbitMQ documentation before going into production.

Having metrics, both of RabbitMQ and applications that use it, is critically important when making informed decisions about production systems.


The example uses, targets or assumes:

Quick Start with Make

This example comes with a Make target that sets up VirtualBox, Minikube and an example cluster in a single command. It can be found under this directory. Homebrew will be used to install packages and on macOS, VirtualBox will need OS permissions to install its kernel module.

The Homebrew cask installer will ask for your password at some point with a prompt that looks like this:

Changing ownership of paths required by virtualbox; your password may be necessary

Please inspect the Make file to be extra sure that you understand and agree to what it does. After enabling 3rd party kernel extensions in OS setings, run the default Make target in this directory:


which is equivalent to first running

make start-minikube

to install VirtualBox and Minikube using Homebrew, then

make run-in-minikube

to start Minikube and kubectl apply the example, and finally

make wait-for-rabbitmq

to wait for cluster formation.

Once the changes are applied, follow the steps in the Check Cluster Status section below.

In case you would prefer to install and run Minikube manually, see the following few sections.

Running the Example Manually with Minikube


  • Make sure that VirtualBox is installed
  • Install minikube and start it with --vm-driver=virtualbox
  • Install kubectl

Start Minikube

Start a minikube virtual machine:

minikube start --cpus=2 --memory=2040 --disk-size "10 GB" --vm-driver=virtualbox

Create a Namespace

Create a Kubernetes namespace for RabbitMQ tests:

kubectl create namespace test-rabbitmq

Set Up Kubernetes Permissions

In Kubernetes 1.6 or above, RBAC authorization is enabled by default. This example configures RBAC related bits so that the peer discovery plugin is allowed to access the nodes information it needs. The ServiceAccount and Role resources will be created in the following step.

kubectl Apply Things

Deploy the config map, services, a stateful set and so on:

# will apply all files under this directory
kubectl create -f examples/minikube

Check Cluster Status

Wait for a a few minutes for pods to start. Since this example uses a stateful set with ordered startup, the pods will be started one by one. To monitor pod startup process, use

kubectl --namespace="test-rabbitmq" get pods

To run rabbitmq-diagnostics cluster_status:

FIRST_POD=$(kubectl get pods --namespace test-rabbitmq -l 'app=rabbitmq' -o jsonpath='{.items[0] }')
kubectl exec --namespace=test-rabbitmq $FIRST_POD rabbitmq-diagnostics cluster_status

to check cluster status. Note that nodes can take some time to start and discover each other.

The output should look something like this:

Cluster status of node rabbit@rabbitmq-0.rabbitmq.test-rabbitmq.svc.cluster.local ...

Cluster name: rabbit@rabbitmq-0.rabbitmq.test-rabbitmq.svc.cluster.local

Disk Nodes


Running Nodes



rabbit@rabbitmq-0.rabbitmq.test-rabbitmq.svc.cluster.local: RabbitMQ 3.8.1 on Erlang 22.1.8
rabbit@rabbitmq-1.rabbitmq.test-rabbitmq.svc.cluster.local: RabbitMQ 3.8.1 on Erlang 22.1.8
rabbit@rabbitmq-2.rabbitmq.test-rabbitmq.svc.cluster.local: RabbitMQ 3.8.1 on Erlang 22.1.8



Network Partitions



Node: rabbit@rabbitmq-0.rabbitmq.test-rabbitmq.svc.cluster.local, interface: [::], port: 25672, protocol: clustering, purpose: inter-node and CLI tool communication
Node: rabbit@rabbitmq-0.rabbitmq.test-rabbitmq.svc.cluster.local, interface: [::], port: 5672, protocol: amqp, purpose: AMQP 0-9-1 and AMQP 1.0
Node: rabbit@rabbitmq-0.rabbitmq.test-rabbitmq.svc.cluster.local, interface: [::], port: 15672, protocol: http, purpose: HTTP API
Node: rabbit@rabbitmq-1.rabbitmq.test-rabbitmq.svc.cluster.local, interface: [::], port: 25672, protocol: clustering, purpose: inter-node and CLI tool communication
Node: rabbit@rabbitmq-1.rabbitmq.test-rabbitmq.svc.cluster.local, interface: [::], port: 5672, protocol: amqp, purpose: AMQP 0-9-1 and AMQP 1.0
Node: rabbit@rabbitmq-1.rabbitmq.test-rabbitmq.svc.cluster.local, interface: [::], port: 15672, protocol: http, purpose: HTTP API

Feature flags

Flag: drop_unroutable_metric, state: enabled
Flag: empty_basic_get_metric, state: enabled
Flag: implicit_default_bindings, state: enabled
Flag: quorum_queue, state: enabled
Flag: virtual_host_metadata, state: enabled

Use Public Minikube IP Address to Connect

Get the public minikube VM IP address:

minikube ip
# =>

The ports used by this example are:

Scaling the Number of RabbitMQ Cluster Nodes (Kubernetes Pod Replicas)

# Odd numbers of nodes are necessary for a clear quorum: 3, 5, 7 and so on
kubectl scale statefulset/rabbitmq --namespace=test-rabbitmq --replicas=5
You can’t perform that action at this time.