Portworx
PORTWORX.COM

RabbitMQ with Portworx on Kubernetes

RabbitMQ is an open-source message broker. It plays a central role in many distributed systems. You can configure RabbitMQ to safely pass messages between decoupled systems.

This reference architecture document shows how you can create and run RabbitMQ with Portworx on Kubernetes. This way, Portworx will provide a reliable persistent storage layer, which makes sure no messages are lost.

The RabbitMQ cluster will use mirrored queues to persist data and metadata to Portworx volumes. Portworx also replicates the volumes, providing multiple layers of redundancy.

Prerequisites

  • You must have a Kubernetes cluster with a minimum of 3 worker nodes.
  • Portworx has been installed on your Kubernetes cluster. For more details on how to install Portworx, refer to the instructions from the Portworx on Kubernetes page.
  • (Optional) Helm must be installed on your Kubernetes cluster if you plan on using it to deploy RabbitMQ. Refer to the Helm Quickstart Guide for details about how to install Helm.

Create a StorageClass

Enter the following kubectl apply command to create a StorageClass:

kubectl apply -f - <<'_EOF'
---
apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  name: portworx-rabbitmq
parameters:
  io_priority: high
  repl: "2"
  group: "rmq_vg"
provisioner: kubernetes.io/portworx-volume
allowVolumeExpansion: true
reclaimPolicy: Delete
_EOF
storageclass.storage.k8s.io/portworx-rabbitmq created

Note the following about this StorageClass:

  • The provisioner parameter is set to kubernetes.io/portworx-volume. For details about the Portworx-specific parameters, refer to the Portworx Volume section.
  • Two replicas of each volume will be created
  • A high-priority storage pool will be used
  • The allowVolumeExpansion parameter is set to true
  • The reclaimPolicy parameter is set to Delete.

You’ll be referencing this StorageClass later in this tutorial when you’ll create a PersistentVolumeClaim. Then, you’ll configure the RabbitMQ Pods to use the persistent volumes created by this PVC.

If you’re using Portworx with CSI, you need to set the value of the provisioner parameter to pxd.portworx.com.

When you install RabbitMQ with Portworx on Kubernetes, you can choose one of the following options:

  • Set up RabbitMQ using Helm
  • Set up RabbitMQ manually

Set up RabbitMQ using Helm

A complex application like RabbitMQ relies on several YAML files to define its various components. This section shows how you can use Helm to simplify the deployment of RabbitMQ.

The following Helm command uses the RabbitMQ High Available Helm chart to create a release named rmq. The release creates a two-member RabbitMQ cluster, and Portworx will mirror queues and messages between the nodes.

helm upgrade \
  --install \
  --set replicaCount=2 \
  --set rabbitmqUsername=admin \
  --set rabbitmqPassword=secretpassword \
  --set managementPassword=anothersecretpassword \
  --set rabbitmqErlangCookie=secretcookie  \
  --set persistentVolume.enabled=true \
  --set persistentVolume.storageClass=portworx-rabbitmq \
  --set schedulerName=stork \
rmq stable/rabbitmq-ha
Release "rmq" does not exist. Installing it now.

NAME: rmq
LAST DEPLOYED: Mon Oct 14 16:58:20 2019
NAMESPACE: default
STATUS: DEPLOYED

Note that the replicaCount variable in this example creates a two-replica StatefulSet and the persistentVolume.storageClass references the StorageClass defined above.

Set up RabbitMQ manually

In this section, we’ll show how you can install RabbitMQ manually.

Note that the spec files used in this section are based on these Helm templates.

  1. Enter the following command to create a ConfigMap and a Kubernetes secret:

    kubectl apply -f - <<'_EOF'
---
# Source: rabbitmq-ha/templates/configmap.yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: rmq-rabbitmq-ha
  namespace: default
  labels:
    app: rabbitmq-ha
data:
  enabled_plugins: |
    [
      rabbitmq_shovel,
      rabbitmq_shovel_management,
      rabbitmq_federation,
      rabbitmq_federation_management,
      rabbitmq_consistent_hash_exchange,
      rabbitmq_management,
      rabbitmq_peer_discovery_k8s
    ].
  rabbitmq.conf: |
    ## RabbitMQ configuration
    ## Ref: https://github.com/rabbitmq/rabbitmq-server/blob/master/docs/rabbitmq.conf.example

    ## Authentification

    ## Clustering
    cluster_formation.peer_discovery_backend  = rabbit_peer_discovery_k8s
    cluster_formation.k8s.host = kubernetes.default.svc.cluster.local
    cluster_formation.k8s.address_type = hostname
    cluster_formation.node_cleanup.interval = 10
    # Set to false if automatic cleanup of absent nodes is desired.
    # This can be dangerous, see http://www.rabbitmq.com/cluster-formation.html#node-health-checks-and-cleanup.
    cluster_formation.node_cleanup.only_log_warning = true
    cluster_partition_handling = autoheal

    ## The default "guest" user is only permitted to access the server
    ## via a loopback interface (e.g. localhost)
    loopback_users.guest = false

    management.load_definitions = /etc/definitions/definitions.json

    ## Memory-based Flow Control threshold
    vm_memory_high_watermark.absolute = 256MB

    ## Auth HTTP Backend Plugin

    ## LDAP Plugin

    ## MQTT Plugin

    ## Web MQTT Plugin

    ## STOMP Plugin

    ## Web STOMP Plugin

    ## Prometheus Plugin

    ## AMQPS support
---
# Source: rabbitmq-ha/templates/secret.yaml
apiVersion: v1
kind: Secret
metadata:
  name: rmq-rabbitmq-ha
  namespace: default
  labels:
    app: rabbitmq-ha
type: Opaque
data:
  rabbitmq-username: "YWRtaW4="
  rabbitmq-password: "c2VjcmV0cGFzc3dvcmQ="
  rabbitmq-management-username: "bWFuYWdlbWVudA=="
  rabbitmq-management-password: "YW5vdGhlcnNlY3JldHBhc3N3b3Jk"
  rabbitmq-erlang-cookie: "c2VjcmV0Y29va2ll"
  definitions.json: "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"
_EOF
    configmap/rmq-rabbitmq-ha created
secret/rmq-rabbitmq-ha created

    Note that we’re separating the application code from the configuration. Refer to the Configuration management with Containers for more details about the ConfigMap resource.

  2. To set up RBAC, apply the following spec in your cluster:

      kubectl apply -f - <<'_EOF'
  ---
  # Source: rabbitmq-ha/templates/serviceaccount.yaml
  apiVersion: v1
  kind: ServiceAccount
  metadata:
    labels:
      app: rabbitmq-ha
    name: rmq-rabbitmq-ha
    namespace: default
  ---
  # Source: rabbitmq-ha/templates/role.yaml
  apiVersion: rbac.authorization.k8s.io/v1
  kind: Role
  metadata:
    labels:
      app: rabbitmq-ha
    name: rmq-rabbitmq-ha
    namespace: default
  rules:
    - apiGroups: [""]
      resources: ["endpoints"]
      verbs: ["get"]
  ---
  # Source: rabbitmq-ha/templates/rolebinding.yaml
  apiVersion: rbac.authorization.k8s.io/v1
  kind: RoleBinding
  metadata:
    labels:
      app: rabbitmq-ha
    name: rmq-rabbitmq-ha
    namespace: default
  subjects:
    - kind: ServiceAccount
      name: rmq-rabbitmq-ha
      namespace: default
  roleRef:
    apiGroup: rbac.authorization.k8s.io
    kind: Role
    name: rmq-rabbitmq-ha
  _EOF
      serviceaccount/rmq-rabbitmq-ha created
  role.rbac.authorization.k8s.io/rmq-rabbitmq-ha created
  rolebinding.rbac.authorization.k8s.io/rmq-rabbitmq-ha created

  3. Enter the following command to create a StatefulSet and the required supporting services:

      kubectl apply -f - <<'_EOF'
  ---
  # Source: rabbitmq-ha/templates/service-discovery.yaml
  apiVersion: v1
  kind: Service
  metadata:
    name: rmq-rabbitmq-ha-discovery
    namespace: default
    labels:
      app: rabbitmq-ha
  spec:
    clusterIP: None
    ports:
      - name: http
        protocol: TCP
        port: 15672
        targetPort: http
      - name: amqp
        protocol: TCP
        port: 5672
        targetPort: amqp
      - name: epmd
        protocol: TCP
        port: 4369
        targetPort: epmd
    publishNotReadyAddresses: true
    selector:
      app: rabbitmq-ha
    type: ClusterIP
  ---
  # Source: rabbitmq-ha/templates/service.yaml
  apiVersion: v1
  kind: Service
  metadata:
    name: rmq-rabbitmq-ha
    namespace: default
    labels:
      app: rabbitmq-ha
  spec:
    clusterIP: "None"
    ports:
      - name: http
        protocol: TCP
        port: 15672
        targetPort: http
      - name: amqp
        protocol: TCP
        port: 5672
        targetPort: amqp
      - name: epmd
        protocol: TCP
        port: 4369
        targetPort: epmd
    selector:
      app: rabbitmq-ha
    type: ClusterIP
  ---
  # Source: rabbitmq-ha/templates/statefulset.yaml
  apiVersion: apps/v1
  kind: StatefulSet
  metadata:
    name: rmq-rabbitmq-ha
    namespace: default
    labels:
      app: rabbitmq-ha
  spec:
    podManagementPolicy: OrderedReady
    serviceName: rmq-rabbitmq-ha-discovery
    replicas: 2
    updateStrategy:
      type: OnDelete
    selector:
      matchLabels:
        app: rabbitmq-ha
    template:
      metadata:
        labels:
          app: rabbitmq-ha
        annotations:
          checksum/config: 1b2c44d6700bf1d0c528a1dc3867f71e5a5e7c9099e15d68639ba407205c1d30
      spec:
        terminationGracePeriodSeconds: 10
        schedulerName: stork
        securityContext:
            fsGroup: 101
            runAsGroup: 101
            runAsNonRoot: true
            runAsUser: 100
        serviceAccountName: rmq-rabbitmq-ha
        initContainers:
          - name: bootstrap
            image: busybox:1.30.1
            imagePullPolicy: IfNotPresent
            command: ['sh']
            args:
            - "-c"
            - |
              set -ex
              cp /configmap/* /etc/rabbitmq
              echo "${RABBITMQ_ERLANG_COOKIE}" > /var/lib/rabbitmq/.erlang.cookie
            env:
            - name: POD_NAME
              valueFrom:
                fieldRef:
                  apiVersion: v1
                  fieldPath: metadata.name
            - name: RABBITMQ_MNESIA_DIR
              value: /var/lib/rabbitmq/mnesia/rabbit@$(POD_NAME).rmq-rabbitmq-ha-discovery.default.svc.cluster.local
            - name: RABBITMQ_ERLANG_COOKIE
              valueFrom:
                secretKeyRef:
                  name: rmq-rabbitmq-ha
                  key: rabbitmq-erlang-cookie
            volumeMounts:
              - name: configmap
                mountPath: /configmap
              - name: config
                mountPath: /etc/rabbitmq
              - name: data
                mountPath: /var/lib/rabbitmq
        containers:
          - name: rabbitmq-ha
            image: rabbitmq:3.7.19-alpine
            imagePullPolicy: IfNotPresent
            ports:
              - name: epmd
                protocol: TCP
                containerPort: 4369
              - name: amqp
                protocol: TCP
                containerPort: 5672
              - name: http
                protocol: TCP
                containerPort: 15672
            livenessProbe:
              exec:
                command:
                - /bin/sh
                - -c
                - 'wget -O - -q --header "Authorization: Basic `echo -n \"$RABBIT_MANAGEMENT_USER:$RABBIT_MANAGEMENT_PASSWORD\"
                  | base64`" http://localhost:15672/api/healthchecks/node | grep -qF "{\"status\":\"ok\"}"'
              failureThreshold: 6
              initialDelaySeconds: 120
              periodSeconds: 10
              timeoutSeconds: 5
            readinessProbe:
              exec:
                command:
                - /bin/sh
                - -c
                - 'wget -O - -q --header "Authorization: Basic `echo -n \"$RABBIT_MANAGEMENT_USER:$RABBIT_MANAGEMENT_PASSWORD\"
                  | base64`" http://localhost:15672/api/healthchecks/node | grep -qF "{\"status\":\"ok\"}"'
              failureThreshold: 6
              initialDelaySeconds: 20
              periodSeconds: 5
              timeoutSeconds: 3
            env:
              - name: MY_POD_NAME
                valueFrom:
                  fieldRef:
                    apiVersion: v1
                    fieldPath: metadata.name
              - name: RABBITMQ_USE_LONGNAME
                value: "true"
              - name: RABBITMQ_NODENAME
                value: rabbit@$(MY_POD_NAME).rmq-rabbitmq-ha-discovery.default.svc.cluster.local
              - name: K8S_HOSTNAME_SUFFIX
                value: .rmq-rabbitmq-ha-discovery.default.svc.cluster.local
              - name: K8S_SERVICE_NAME
                value: rmq-rabbitmq-ha-discovery
              - name: RABBITMQ_ERLANG_COOKIE
                valueFrom:
                  secretKeyRef:
                    name: rmq-rabbitmq-ha
                    key: rabbitmq-erlang-cookie
              - name: RABBIT_MANAGEMENT_USER
                valueFrom:
                  secretKeyRef:
                    name: rmq-rabbitmq-ha
                    key: rabbitmq-management-username
              - name: RABBIT_MANAGEMENT_PASSWORD
                valueFrom:
                  secretKeyRef:
                    name: rmq-rabbitmq-ha
                    key: rabbitmq-management-password
            volumeMounts:
              - name: data
                mountPath: /var/lib/rabbitmq
              - name: config
                mountPath: /etc/rabbitmq
              - name: definitions
                mountPath: /etc/definitions
                readOnly: true
        affinity:
          podAntiAffinity:
            preferredDuringSchedulingIgnoredDuringExecution:
              - weight: 1
                podAffinityTerm:
                  topologyKey: "kubernetes.io/hostname"
                  labelSelector:
                    matchLabels:
                      app: rabbitmq-ha
        volumes:
          - name: config
            emptyDir: {}
          - name: configmap
            configMap:
              name: rmq-rabbitmq-ha
          - name: definitions
            secret:
              secretName: rmq-rabbitmq-ha
              items:
              - key: definitions.json
                path: definitions.json
    volumeClaimTemplates:
      - metadata:
          name: data
          annotations:
        spec:
          accessModes:
            - "ReadWriteOnce"
          resources:
            requests:
              storage: "8Gi"
          storageClassName: "portworx-rabbitmq"
  _EOF
      service/rmq-rabbitmq-ha-discovery created
  service/rmq-rabbitmq-ha created
  statefulset.apps/rmq-rabbitmq-ha created

    Note that the last line of the spec references the portworx-rabbitmq storage class defined in the Create a StorageClass section. As a result, Kubernetes will automatically create a new PVC for each replica.

Validate the cluster functionality

Once you’ve set up RabbitMQ manually or using Helm, you can test your installation using a tool called PerfTest. In the next sections, we will use PerfTest to:

  • simulate basic workloads
  • measure the performance of the system
  • introduce a failure and see how the cluster recovers from it.

Create a RabbitMQ policy for H/A

Configure a policy named perf-test-with-ha with the following settings:

  • The rabbitmqctl set_policy perf-test-with-ha '^perf-test' portion of the command only applies these settings to queues that begin with perf-test
  • Sets the ha-mode parameter equal to 2.
  • Places the queue leader on the least-loaded node.

  • Sets up the queues as lazy. RabbitMQ will save the queues to disk as early as possible.

    kubectl exec rmq-rabbitmq-ha-0 -- \
rabbitmqctl set_policy perf-test-with-ha '^perf-test' \
'{
"ha-mode":"exactly",
"ha-params":2,
"ha-sync-mode":"automatic",
"queue-master-locator":"min-masters",
"queue-mode":"lazy"
}' --apply-to queues
    Setting policy "perf-test-with-ha" for pattern "^perf-test" to "{
"ha-mode":"exactly",
"ha-params":2,
"ha-sync-mode":"automatic",
"queue-master-locator":"min-masters",
"queue-mode":"lazy"
}" with priority "0" for vhost "/" ...

Set up monitoring

RabbitMQ provides a web-based user interface that you can use to manage your cluster. To access it, you must forward connections made to a local port to the 15672 port on the rmq-rabbitmq-ha-0 Pod.

  1. Enter the following kubectl port-forward command to forward all connections made to localhost:15672 to rmq-rabbitmq-ha-0:15672:

      kubectl port-forward rmq-rabbitmq-ha-0 15672:15672
      Forwarding from 127.0.0.1:15672 -> 15672

  2. You can now access the RabbitMQ user interface by pointing your browser to localhost:1567. Use the credentials created in the Set up RabbitMQ using Helm section.

  3. Visit the Queue section at your localhost: http://127.0.0.1:15672/#/queues, to see that the perf-test-with-ha policy is applied.

  4. Additionally, you can stream to stdout the logs of the Pod that survives the failover test by entering the following command:

      kubectl logs -f rmq-rabbitmq-ha-0

Create a containerized testing environment

The following kubectl run command creates a Pod with a sleeping container that runs for an hour (see the "args": ["-c", "sleep 3600"] line). Then, Kubernetes automatically restarts the container (see the --restart=Always line).

kubectl run perftest \
  --restart=Always \
  --image dummy-required-param \
  --generator=run-pod/v1 \
  --overrides \
'{
  "spec": {
    "affinity": {
      "podAntiAffinity": {
        "requiredDuringSchedulingIgnoredDuringExecution": [{
          "labelSelector": {
            "matchExpressions": [{
              "key": "app",
              "operator": "In",
              "values": ["rabbitmq-ha"]
            }]
          },
          "topologyKey": "kubernetes.io/hostname"
        }]
      }
    },
    "containers": [{
      "command": ["sh"],
      "image": "pivotalrabbitmq/perf-test:latest",
      "name": "perftest",
      "args": ["-c", "sleep 3600"]
    }],
    "securityContext": {
      "runAsUser": 1000
    }
  }
}'
pod/perftest created
Note that the above command creates a Pod with one sleeping container. The Pod is then scheduled onto a node that doesn’t run RabbitMQ. This way, you evenly distribute bandwidth between both RabbitMQ nodes. Also note that we use a non-root user to run the commands. This is useful for when you have PodSecurityPolicies enabled.

Simulate basic workloads

To start the simulated producers and consumers, run the bin/runjava command in the perftest container:

kubectl exec -it perftest -- \
    bin/runjava com.rabbitmq.perf.PerfTest \
      --time 900 \
      --queue-pattern 'perf-test-%d' \
      --queue-pattern-from 1 \
      --queue-pattern-to 2 \
      --producers 2 \
      --consumers 8 \
      --queue-args x-cancel-on-ha-failover=true \
      --flag persistent \
      --uri amqp://admin:secretpassword@rmq-rabbitmq-ha:5672?failover=failover_exchange

There are several things to note about this example:

  • Runs the session for 15 minutes with queues named in the format of perf-test- followed by a number (in our case 1 and 2)
  • Creates threads for 2 producers and 8 consumers (so theoretically each RabbitMQ replica hosts 1 queue and gets 1 producer and 4 consumers
  • Trades off performance for reliability, by flagging the messages as persistent

Failover

Simulate a failure by force deleting the rmq-rabbitmq-ha-1 Pod:

kubectl delete pod rmq-rabbitmq-ha-1 --force --grace-period=0
warning: Immediate deletion does not wait for confirmation that the running resource has been terminated. The resource may continue to run on the cluster indefinitely.
pod "rmq-rabbitmq-ha-1" force deleted

As the system recovers from the failure, you will see that the performance testing session will pause for a short period. As a result, the messages will pile up in the queue. After the failover, the cluster resumes normal processing, and no messages are lost.

Clean up

If you used Helm to install the RabbitMQ cluster, then enter the following command to remove all the Kubernetes components and delete the release:

helm delete --purge rmq

If you installed the RabbitMQ cluster manually, then enter the following command to delete all the associated resources:

kubectl delete sts/rmq-rabbitmq-ha \
  svc/rmq-rabbitmq-ha svc/rmq-rabbitmq-ha-discovery \
  rolebinding/rmq-rabbitmq-ha \
  role/rmq-rabbitmq-ha \
  sa/rmq-rabbitmq-ha \
  secret/rmq-rabbitmq-ha \
  cm/rmq-rabbitmq-ha

Finally, regardless of the deployment method, execute the following command:

kubectl delete pvc/data-rmq-rabbitmq-ha-0 pvc/data-rmq-rabbitmq-ha-1 \
  sc/portworx-rabbitmq \
  po/perftest

These kubectl delete commands remove the following:

  • The workload itself (StatefulSet and the associated services)
  • The RBAC settings for the workload (Rolebinding, Role, and ServiceAccount)
  • The RBAC configuration (ConfigMap and secret)
  • The volumes including data (PersistentVolumeClaims)
  • The volume parameters (StorageClass)
  • The Pod we used for testing
Last edited: Tuesday, May 9, 2023
Questions? Visit the Portworx forum.