RabbitMQ Cluster Operator (Production)

For production deployments, it is strongly recommended to use the official RabbitMQ Cluster Kubernetes Operator instead of the Bitnami Helm chart. The operator provides better lifecycle management, high availability, and production-grade features.

The bundled Bitnami subchart is stuck on RabbitMQ 4.1 and cannot be upgraded. The chart’s rabbitmq dependency pulls bitnamilegacy/rabbitmq, a registry that’s frozen at 4.1.3-debian-12-r1 — no 4.2.x/4.3.x tags exist (confirmed via docker manifest inspect against several candidate tags on 2026-08-03, and reproduced in-cluster: helm install with a bumped rabbitmq.image.tag leaves the RabbitMQ pod in ImagePullBackOff/ ErrImagePull — ... not found). Since RabbitMQ 4.1 loses community support in January 2026 and 4.2 reaches EOL July 31, 2026, the bundled subchart cannot be the long-term path for any deployment. This is why waldur/test/values.yaml (CI’s test install target) now points at an operator-managed cluster instead of the subchart. Migrating the bundled subchart itself off the frozen registry — to a different chart/image entirely — is tracked as separate future work, not part of this compatibility fix.

Overview

The RabbitMQ Cluster Operator automates:

Provisioning and management of RabbitMQ clusters
Scaling and automated rolling upgrades
Monitoring integration with Prometheus and Grafana
Backup and recovery operations
Network policy and security configurations

Prerequisites

Kubernetes cluster version 1.19 or above
Configured kubectl access
Appropriate RBAC permissions

Version upgrade path

RabbitMQ does not support skipping a minor version. From the upstream upgrade guide: “Upgrading to 4.3.x is only possible from 4.2.x.” The supported steps are 4.1.x → 4.2.x and 4.2.x → 4.3.x; there is no 4.1.x → 4.3.x path. A cluster on 4.1 must therefore be walked through 4.2, rolling and verifying at each step, rather than jumped straight to 4.3.

Before each step:

Enable all stable feature flags. The upgrade guide is explicit that “all stable feature flags must be enabled before an upgrade, or the upgrade may fail”. This is not hypothetical — a default 4.1 node has khepri_db disabled, and 4.3 requires it.
```
kubectl exec -n <namespace> <cluster-name>-server-0 -- rabbitmqctl enable_feature_flag all
kubectl exec -n <namespace> <cluster-name>-server-0 -- rabbitmqctl list_feature_flags
```
Check the Erlang requirement. RabbitMQ 4.3.3 and later require Erlang 27, and nodes fail to start on anything older. The official rabbitmq:4.3.x container images already ship OTP 27, so this only affects package-based installations.

Bump spec.image one minor series at a time and let the operator roll the StatefulSet, waiting for AllReplicasReady before starting the next step:

kubectl wait --for=condition=AllReplicasReady --timeout=600s rabbitmqcluster/<cluster-name>

Installation

1. Install the RabbitMQ Cluster Operator

kubectl apply -f https://github.com/rabbitmq/cluster-operator/releases/download/v2.22.5/cluster-operator.yml

Pin the version rather than using latest: the operator is cluster-scoped, so an unpinned re-apply upgrades it for every RabbitmqCluster on the cluster at once.

Verify the operator is running:

kubectl get pods -n rabbitmq-system

2. Create a Production RabbitMQ Cluster

Create a production-ready RabbitMQ cluster configuration:

apiVersion: rabbitmq.com/v1beta1
kind: RabbitmqCluster
metadata:
  name: waldur-rabbitmq
  namespace: default
spec:
  # Pin explicitly -- an unset image field silently takes whatever the
  # operator's default is, rather than a deliberate version.
  image: rabbitmq:4.3.5

  # Production recommendation: use odd numbers (3, 5, 7)
  replicas: 3

  # Resource configuration
  resources:
    requests:
      cpu: 1000m      # 1 CPU core
      memory: 2Gi     # Keep requests and limits equal for stability
    limits:
      cpu: 2000m      # 2 CPU cores for peak loads
      memory: 2Gi

  # Persistence configuration
  persistence:
    storageClassName: "fast-ssd"  # Use appropriate storage class
    storage: 20Gi                 # Adjust based on expected message volume

  # RabbitMQ configuration
  rabbitmq:
    additionalConfig: |
      # Memory threshold (80% of available memory)
      vm_memory_high_watermark.relative = 0.8

      # Disk threshold (2GB free space)
      disk_free_limit.absolute = 2GB

      # Clustering settings
      cluster_formation.peer_discovery_backend = rabbit_peer_discovery_k8s
      cluster_formation.k8s.host = kubernetes.default.svc.cluster.local
      cluster_formation.node_cleanup.interval = 30
      cluster_formation.node_cleanup.only_log_warning = true

      # Management plugin
      management.tcp.port = 15672

      # Enable additional protocols if needed
      listeners.tcp.default = 5672

      # Logging
      log.console = true
      log.console.level = info

      # Queue leader location policy (queue_master_locator is deprecated)
      queue_leader_locator = balanced

      # RabbitMQ 4.3 compatibility: 4.3 denies transient, non-exclusive
      # queues by default, the combination kombu's pidbox and Celery's
      # gossip queues declare. Mastermind now declares both exclusive
      # (CELERY_CONTROL_QUEUE_EXCLUSIVE / CELERY_EVENT_QUEUE_EXCLUSIVE), so
      # this flag is only needed while the deployed Waldur image predates
      # that change. It is read at node boot and is accepted on 4.1, 4.2 and
      # 4.3 alike, so it can be set or removed at any point in the walk.
      deprecated_features.permit.transient_nonexcl_queues = true

    # Additional plugins
    additionalPlugins:

      - rabbitmq_management

      - rabbitmq_prometheus

      - rabbitmq_auth_backend_ldap  # If LDAP auth is needed

      - rabbitmq_stomp              # If STOMP protocol is needed

      - rabbitmq_web_stomp          # Required for Waldur's websocket-STOMP ingress

  # Service configuration
  service:
    type: ClusterIP
    annotations:
      service.beta.kubernetes.io/aws-load-balancer-type: nlb  # For AWS

  # Monitoring
  override:
    statefulSet:
      spec:
        template:
          metadata:
            annotations:
              prometheus.io/scrape: "true"
              prometheus.io/port: "15692"
              prometheus.io/path: "/metrics"

  # Security and networking
  affinity:
    podAntiAffinity:
      preferredDuringSchedulingIgnoredDuringExecution:

      - weight: 100

        podAffinityTerm:
          labelSelector:
            matchExpressions:

            - key: app.kubernetes.io/name

              operator: In
              values:

              - rabbitmq

          topologyKey: kubernetes.io/hostname

Apply the configuration:

kubectl apply -f waldur-rabbitmq-cluster.yaml

Configuration for Waldur

1. Retrieve RabbitMQ Credentials

Get the auto-generated credentials:

# Get username

kubectl get secret waldur-rabbitmq-default-user -o jsonpath='{.data.username}' | base64 --decode

# Get password

kubectl get secret waldur-rabbitmq-default-user -o jsonpath='{.data.password}' | base64 --decode

2. Configure Waldur Helm Values

Update your Waldur values.yaml:

# Disable the bundled rabbitmq subchart and point the chart at the
# operator-managed cluster. `host` must match the RabbitmqCluster name -- the
# operator creates a Service of that name -- and the secret is the one the
# operator generates as `<cluster-name>-default-user`.

rabbitmq:
  enabled: false
  host: "waldur-rabbitmq"
  secret:
    name: "waldur-rabbitmq-default-user"
    usernameKey: "username"
    passwordKey: "password"

Do not set global.waldur.rabbitmq.*. The chart does not read those keys; the broker host comes from the top-level rabbitmq.host value (templates/_helpers.tpl, waldur.rabbitmq.rmqHost). Setting only the global.* form leaves rabbitmq.host at its default of rmq-rabbitmq — the name of the old Bitnami release — and the deployment points at a Service that does not exist.

RabbitMQ Operator Secret Management:

The RabbitMQ Cluster Operator automatically creates a default user secret named [cluster-name]-default-user containing:

username - Auto-generated username
password - Auto-generated password
Other connection details

This approach avoids hardcoding credentials and follows Kubernetes security best practices.

High Availability Configuration

For production high availability, consider these additional configurations:

Pod Disruption Budget

apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: waldur-rabbitmq-pdb
spec:
  minAvailable: 2  # Ensure at least 2 pods are always available
  selector:
    matchLabels:
      app.kubernetes.io/name: waldur-rabbitmq

Network Policy (Optional)

Restrict network access to RabbitMQ:

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: waldur-rabbitmq-netpol
spec:
  podSelector:
    matchLabels:
      app.kubernetes.io/name: waldur-rabbitmq
  policyTypes:

  - Ingress

  ingress:

  - from:

    - podSelector:

        matchLabels:
          app.kubernetes.io/name: waldur
    ports:

    - protocol: TCP

      port: 5672

  - from:  # Allow management interface access

    - podSelector:

        matchLabels:
          app: monitoring
    ports:

    - protocol: TCP

      port: 15672

    - protocol: TCP

      port: 15692  # Prometheus metrics

Monitoring

The operator automatically enables Prometheus metrics. To access them:

Prometheus Metrics Endpoint: <http://waldur-rabbitmq:15692/metrics>

Management UI Access:

kubectl port-forward service/waldur-rabbitmq 15672:15672

Access at: <http://localhost:15672>

Grafana Dashboard: Import RabbitMQ dashboard ID 10991 or similar

Backup and Recovery

Automated Backup Configuration

The operator supports backup configurations through definitions:

apiVersion: rabbitmq.com/v1beta1
kind: Backup
metadata:
  name: waldur-rabbitmq-backup
spec:
  rabbitmqClusterReference:
    name: waldur-rabbitmq

For production, implement external backup strategies using tools like Velero or cloud-native backup solutions.

Scaling

Scale the cluster:

kubectl patch rabbitmqcluster waldur-rabbitmq --type='merge' -p='{"spec":{"replicas":5}}'

Important: Always use odd numbers for replicas (1, 3, 5, 7) to avoid split-brain scenarios.

Troubleshooting

Check Cluster Status

# Check pods

kubectl get pods -l app.kubernetes.io/name=waldur-rabbitmq

# Check cluster status

kubectl exec waldur-rabbitmq-server-0 -- rabbitmq-diagnostics cluster_status

# Check node health

kubectl exec waldur-rabbitmq-server-0 -- rabbitmq-diagnostics check_running

View Logs

# View operator logs

kubectl logs -n rabbitmq-system deployment/rabbitmq-cluster-operator

# View RabbitMQ logs

kubectl logs waldur-rabbitmq-server-0

Migration from Bitnami Chart

If migrating from the Bitnami chart:

Backup existing data using RabbitMQ management tools
Deploy the operator and create a new cluster
Export/import virtual hosts, users, and permissions
Update Waldur configuration to point to the new cluster
Test thoroughly before decommissioning the old setup

Security Considerations

TLS Configuration: Enable TLS for production:

spec:
  tls:
    secretName: waldur-rabbitmq-tls

Authentication: Consider integrating with LDAP or other authentication backends
Network Policies: Implement network policies to restrict access
RBAC: Ensure appropriate Kubernetes RBAC policies are in place

Performance Tuning

For high-throughput scenarios:

Adjust memory limits based on message volume
Configure disk I/O with appropriate storage classes
Tune RabbitMQ parameters in additionalConfig
Monitor resource usage and scale accordingly

Support and Documentation

Official Documentation: https://www.rabbitmq.com/kubernetes/operator/
GitHub Repository: https://github.com/rabbitmq/cluster-operator
Examples: https://github.com/rabbitmq/cluster-operator/tree/main/docs/examples
Community Support: RabbitMQ Discussions on GitHub

waldur-helm

Packaging of Waldur as Helm application