Deploy Microservices On-Premises với Kubernetes HA

Phần 5: Message Queue HA (RabbitMQ, Kafka, Redis)

xdev.asia

🎯 MỤC TIÊU BÀI HỌC

• ✅ Hiểu kiến trúc RabbitMQ cluster và các messaging patterns
• ✅ Deploy RabbitMQ 3-node cluster bằng Cluster Operator
• ✅ Cấu hình quorum queues cho HA
• ✅ Setup TLS encryption và authentication
• ✅ Monitor RabbitMQ cluster với Prometheus
• ✅ Best practices: message durability, DLQ, flow control

---

PHẦN 1: KIẾN TRÚC RABBITMQ CLUSTER

1.1. Messaging Patterns

graph LR
    subgraph P2P["1️⃣ Point-to-Point"]
        PP["Producer"] --> Q1["Queue"] --> C1A["Consumer"]
        Q1 --> C1B["Consumer"]
    end

    subgraph FANOUT["2️⃣ Pub/Sub Fanout"]
        FP["Producer"] --> EX1["Exchange"]
        EX1 --> FQ1["Queue1"] --> FC1["Consumer1"]
        EX1 --> FQ2["Queue2"] --> FC2["Consumer2"]
    end

    subgraph ROUTING["3️⃣ Routing"]
        RP["Producer"] -->|"routing.key"| EX2["Exchange"]
        EX2 -->|"match"| RQ["Queue"] --> RC["Consumer"]
    end

    style P2P fill:#0f172a,stroke:#3b82f6,color:#e2e8f0
    style FANOUT fill:#0f172a,stroke:#7c3aed,color:#e2e8f0
    style ROUTING fill:#0f172a,stroke:#15803d,color:#e2e8f0

1.2. RabbitMQ Cluster Architecture

graph TD
    SVC["🌐 K8s Service ClusterIP<br/>rabbitmq.messaging:5672"]

    SVC --> R0
    SVC --> R1
    SVC --> R2

    subgraph CLUSTER["🐰 RabbitMQ Quorum Cluster"]
        R0["🟢 rabbit-0<br/>Leader<br/>Mnesia"]
        R1["🔵 rabbit-1<br/>Follower<br/>Mnesia"]
        R2["🔵 rabbit-2<br/>Follower<br/>Mnesia"]
        R0 <-->|"Raft replication"| R1
        R1 <-->|"Raft replication"| R2
    end

    R0 --> PVC0["💾 Ceph PVC 20Gi"]
    R1 --> PVC1["💾 Ceph PVC 20Gi"]
    R2 --> PVC2["💾 Ceph PVC 20Gi"]

    style SVC fill:#7c3aed,stroke:#a78bfa,color:#e2e8f0
    style CLUSTER fill:#0f172a,stroke:#3b82f6,color:#e2e8f0
    style R0 fill:#15803d,stroke:#22c55e,color:#e2e8f0
    style R1 fill:#1e3a5f,stroke:#3b82f6,color:#e2e8f0
    style R2 fill:#1e3a5f,stroke:#3b82f6,color:#e2e8f0
    style PVC0 fill:#1e293b,stroke:#475569,color:#e2e8f0
    style PVC1 fill:#1e293b,stroke:#475569,color:#e2e8f0
    style PVC2 fill:#1e293b,stroke:#475569,color:#e2e8f0

Quorum Queue: Raft consensus → data replicated across majority Classic Queue: Only on 1 node (mirrored = deprecated)

Feature	Classic Queue	Quorum Queue	Stream
Replication	None (mirrored deprecated)	Raft-based (majority)	Append-only log
Data Safety	Low	High	High
Performance	Highest	Good (slightly lower)	Best for fan-out
Use Case	Temp/non-critical	Business-critical	Event streaming
Ordering	FIFO	FIFO	FIFO per partition

---

PHẦN 2: CÀI ĐẶT RABBITMQ CLUSTER OPERATOR

2.1. Install Operator

# Install RabbitMQ Cluster Operator:
kubectl apply -f https://github.com/rabbitmq/cluster-operator/releases/latest/download/cluster-operator.yml
Verify:
kubectl get pods -n rabbitmq-system
NAME                                         READY   STATUS
rabbitmq-cluster-operator-7f7d8b8bb9-xxxxx   1/1     Running
CRDs installed:
kubectl get crds | grep rabbitmq
rabbitmqclusters.rabbitmq.com

2.2. Tạo Namespace cho Messaging

kubectl create namespace messaging
kubectl label namespace messaging purpose=message-brokers

---

PHẦN 3: DEPLOY RABBITMQ HA CLUSTER

3.1. RabbitmqCluster CRD

# rabbitmq-cluster.yaml:
apiVersion: rabbitmq.com/v1beta1
kind: RabbitmqCluster
metadata:
  name: production-rmq
  namespace: messaging
spec:
  replicas: 3
image: rabbitmq:3.13-management
resources:
requests:
cpu: "500m"
memory: "1Gi"
limits:
cpu: "2"
memory: "2Gi"
persistence:
storageClassName: ceph-block
storage: 20Gi
rabbitmq:
additionalConfig: |
# Cluster formation
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
cluster_formation.node_cleanup.only_log_warning = true
cluster_partition_handling = pause_minority
  # Quorum queues as default:
  default_queue_type = quorum

  # Memory management:
  vm_memory_high_watermark.relative = 0.7
  vm_memory_high_watermark_paging_ratio = 0.8
  disk_free_limit.relative = 1.5

  # Connection limits:
  channel_max = 2047
  heartbeat = 60

  # Consumer timeout (prevent stuck consumers):
  consumer_timeout = 3600000

  # Management plugin stats collection:
  collect_statistics_interval = 10000

advancedConfig: |
  [
    {rabbit, [
      {tcp_listen_options, [
        {backlog, 128},
        {nodelay, true},
        {linger, {true, 0}},
        {exit_on_close, false}
      ]}
    ]}
  ].

additionalPlugins:
  - rabbitmq_prometheus
  - rabbitmq_shovel
  - rabbitmq_shovel_management

affinity:
podAntiAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchLabels:
app.kubernetes.io/name: production-rmq
topologyKey: kubernetes.io/hostname
override:
statefulSet:
spec:
template:
spec:
topologySpreadConstraints:
- maxSkew: 1
topologyKey: kubernetes.io/hostname
whenUnsatisfiable: DoNotSchedule
labelSelector:
matchLabels:
app.kubernetes.io/name: production-rmq

# Deploy:
kubectl apply -f rabbitmq-cluster.yaml

# Watch pods:
kubectl -n messaging get pods -w
# production-rmq-server-0   1/1   Running   0   60s
# production-rmq-server-1   1/1   Running   0   90s
# production-rmq-server-2   1/1   Running   0   120s

# Verify cluster:
kubectl -n messaging exec production-rmq-server-0 -- rabbitmqctl cluster_status

# Services created:
kubectl -n messaging get svc
# production-rmq            ClusterIP   10.96.x.x   5672,15672,15692
# production-rmq-nodes      ClusterIP   None        4369,25672

---

PHẦN 4: QUORUM QUEUES VÀ POLICIES

4.1. Tạo Quorum Queues

# Lấy credentials:
kubectl -n messaging get secret production-rmq-default-user -o jsonpath='{.data.username}' | base64 -d
kubectl -n messaging get secret production-rmq-default-user -o jsonpath='{.data.password}' | base64 -d
Port-forward management UI:
kubectl -n messaging port-forward svc/production-rmq 15672:15672
Via CLI - tạo quorum queue:
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqadmin declare queue name=orders.created 

queue_type=quorum durable=true 

arguments='{"x-quorum-initial-group-size": 3, "x-delivery-limit": 5}'
Queue với Dead Letter Exchange:
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqadmin declare exchange name=orders type=topic durable=true
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqadmin declare exchange name=orders.dlx type=fanout durable=true
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqadmin declare queue name=orders.dlq 

queue_type=quorum durable=true
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqadmin declare binding source=orders.dlx 

destination=orders.dlq destination_type=queue
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqadmin declare queue name=orders.process 

queue_type=quorum durable=true 

arguments='{"x-dead-letter-exchange": "orders.dlx", "x-delivery-limit": 3}'

4.2. Virtual Hosts & Users

# Create vhost cho mỗi service:
kubectl -n messaging exec production-rmq-server-0 -- \
  rabbitmqctl add_vhost /orders
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqctl add_vhost /payments
Create service users:
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqctl add_user order_service "$(openssl rand -base64 24)"
kubectl -n messaging exec production-rmq-server-0 -- 

rabbitmqctl set_permissions -p /orders order_service "orders\.." "orders\.." "orders\..*"
Giới hạn permissions (principle of least privilege):
configure: "orders\..*"   → chỉ configure queues prefix orders.
write:     "orders\..*"   → chỉ publish vào exchanges prefix orders.
read:      "orders\..*"   → chỉ consume từ queues prefix orders.

---

PHẦN 5: TLS ENCRYPTION

# Sử dụng cert-manager:
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
  name: rabbitmq-tls
  namespace: messaging
spec:
  secretName: rabbitmq-tls-secret
  issuerRef:
    name: cluster-ca-issuer
    kind: ClusterIssuer
  commonName: production-rmq.messaging.svc
  dnsNames:
    - production-rmq.messaging.svc
    - production-rmq.messaging.svc.cluster.local
    - "*.production-rmq-nodes.messaging.svc.cluster.local"
---
# Update RabbitmqCluster:
apiVersion: rabbitmq.com/v1beta1
kind: RabbitmqCluster
metadata:
  name: production-rmq
  namespace: messaging
spec:
  tls:
    secretName: rabbitmq-tls-secret
    caSecretName: rabbitmq-ca-secret
    disableNonTLSListeners: true   # ⚠️ Force TLS only

---

PHẦN 6: MONITORING

# PodMonitor cho Prometheus:
apiVersion: monitoring.coreos.com/v1
kind: PodMonitor
metadata:
  name: rabbitmq-monitor
  namespace: messaging
spec:
  selector:
    matchLabels:
      app.kubernetes.io/name: production-rmq
  podMetricsEndpoints:
    - port: prometheus
      interval: 15s
      path: /metrics
---
# Key metrics:
# rabbitmq_queue_messages_ready         — Messages waiting
# rabbitmq_queue_messages_unacked       — Messages in-flight
# rabbitmq_queue_consumers              — Consumer count
# rabbitmq_connections                  — Total connections
# rabbitmq_channels                     — Total channels
# rabbitmq_process_resident_memory_bytes — RAM usage
# rabbitmq_disk_space_available_bytes   — Disk available

# Alerting rules:
apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: rabbitmq-alerts
  namespace: messaging
spec:
  groups:
    - name: rabbitmq
      rules:
        - alert: RabbitMQQueueBacklog
          expr: rabbitmq_queue_messages_ready > 10000
          for: 5m
          labels:
            severity: warning
          annotations:
            summary: "Queue {{ $labels.queue }} has {{ $value }} messages"

        - alert: RabbitMQNoConsumers
          expr: rabbitmq_queue_consumers == 0
          for: 10m
          labels:
            severity: critical
          annotations:
            summary: "Queue {{ $labels.queue }} has no consumers"

        - alert: RabbitMQHighMemory
          expr: rabbitmq_process_resident_memory_bytes / rabbitmq_resident_memory_limit_bytes > 0.8
          for: 5m
          labels:
            severity: warning

        - alert: RabbitMQNodeDown
          expr: rabbitmq_identity_info < 3
          for: 1m
          labels:
            severity: critical

---

PHẦN 7: APPLICATION INTEGRATION

# Application sử dụng RabbitMQ:
apiVersion: apps/v1
kind: Deployment
metadata:
  name: order-service
  namespace: default
spec:
  template:
    spec:
      containers:
        - name: order-service
          env:
            - name: RABBITMQ_URL
              value: "amqps://order_service:$(RABBITMQ_PASSWORD)@production-rmq.messaging:5671/orders"
            - name: RABBITMQ_PASSWORD
              valueFrom:
                secretKeyRef:
                  name: order-rmq-secret
                  key: password

# Python producer (pika):
import pika
import json

connection = pika.BlockingConnection(
    pika.URLParameters(os.environ['RABBITMQ_URL'])
)
channel = connection.channel()

# Publish with delivery confirmation:
channel.confirm_delivery()

message = {"order_id": "12345", "amount": 100.00}
channel.basic_publish(
    exchange='orders',
    routing_key='orders.created',
    body=json.dumps(message),
    properties=pika.BasicProperties(
        delivery_mode=2,        # Persistent message
        content_type='application/json',
        message_id=str(uuid.uuid4())
    )
)

---

💡 KEY TAKEAWAYS

1. Quorum queues: Raft-based replication, always use for production
2. Cluster Operator: Declarative RabbitMQ on K8s, auto-clustering
3. 3-node cluster: Tolerates 1 node failure (majority = 2)
4. Dead Letter Queue: Handle poison messages, retry logic
5. TLS + vhosts: Isolate services, encrypt in transit
6. pause_minority: Prevent split-brain partition handling

---

🎯 BÀI TẬP

Bài tập 1: RabbitMQ HA Lab

• Deploy 3-node RabbitMQ cluster
• Create quorum queue with DLQ
• Publish 10,000 messages, kill 1 node, verify no message loss

Bài tập 2: Monitoring Setup

• Configure PodMonitor
• Import RabbitMQ Grafana dashboard (ID: 10991)
• Create alert for queue backlog > 5000

---

📚 BÀI TIẾP THEO

Trong Bài 22: Apache Kafka Cluster với Strimzi Operator, chúng ta sẽ deploy Kafka cho high-throughput event streaming.