Docker Swarm上基于Consul实现高可用RabbitMQ集群

本文实验通过Docker swarm建立基于Consul自动发现的高可用性的RabbitMQ集群[1]。我们还将添加HAProxy服务器以负载均衡AMQP请求，并提高群集的整体可用性。以下是最终的逻辑拓扑。

安装docker swarm集群

 

主机名	IP
master	192.168.99.128
Node1	192.168.99.133
Node2	192.168.99.134

 

安装docker

curl -fsSL get.docker.com -o get-docker.sh
CHANNEL=stable sh get-docker.sh --mirror Aliyun
root@master:~# docker -v
Docker version 19.03.12, build 48a66213fe

初始化docker swarm集群

• 在master上初始化集群

#Init Docker Swarm - run on master
root@master:~# docker swarm init --advertise-addr 192.168.99.128
Swarm initialized: current node (htu24tq7mhip4ca7a9uqnndef) is now a manager.

To add a worker to this swarm, run the following command:

    docker swarm join --token SWMTKN-1-1ifiwgcc2k4ztyr80zm6e3vj7ewbbi1fqfnba1vfisesnos6o4-es73sy0dx4ie5lbivx90mcbky 192.168.99.128:2377

To add a manager to this swarm, run 'docker swarm join-token manager' and follow the instructions.

• node1加入集群

root@node1:~# docker swarm join --token SWMTKN-1-1ifiwgcc2k4ztyr80zm6e3vj7ewbbi1fqfnba1vfisesnos6o4-es73sy0dx4ie5lbivx90mcbky 192.168.99.128:2377
This node joined a swarm as a worker.

root@master:~# docker node ls
ID                            HOSTNAME              STATUS              AVAILABILITY        MANAGER STATUS      ENGINE VERSION
htu24tq7mhip4ca7a9uqnndef *   master.devopsman.cn   Ready               Active              Leader              19.03.12
vlfb0ibx4k1z41r76yhgpummk     node1.devopsman.cn    Ready               Active                                  19.03.12
root@master:~# docker node promote node1.devopsman.cn
Node node1.devopsman.cn promoted to a manager in the swarm.

• node2加入集群

root@node2:~# docker swarm join --token SWMTKN-1-1ifiwgcc2k4ztyr80zm6e3vj7ewbbi1fqfnba1vfisesnos6o4-es73sy0dx4ie5lbivx90mcbky 192.168.99.128:2377
This node joined a swarm as a worker.
root@master:~# docker node ls
ID                            HOSTNAME              STATUS              AVAILABILITY        MANAGER STATUS      ENGINE VERSION
htu24tq7mhip4ca7a9uqnndef *   master.devopsman.cn   Ready               Active              Leader              19.03.12
vlfb0ibx4k1z41r76yhgpummk     node1.devopsman.cn    Ready               Active              Reachable           19.03.12
99f5xkfepct301n48tqqslz22     node2.devopsman.cn    Ready               Active                                  19.03.12
root@master:~# docker node promote node2.devopsman.cn
Node node2.devopsman.cn promoted to a manager in the swarm.

root@master:~# docker network create --driver=overlay --attachable prod
q5rzw0i1xdh9kovytjeg0sj9p

root@master:~# docker network ls
NETWORK ID          NAME                            DRIVER              SCOPE
ba6f05d438cf        bridge                          bridge              local
ee47cd6a4247        docker_gwbridge                 bridge              local
8df9d4ecb5be        host                            host                local
rl5o9jz6fxp8        ingress                         overlay             swarm
25lk998untvk        local                           overlay             swarm
3a66aac75dbf        none                            null                local
q5rzw0i1xdh9        prod                            overlay             swarm

部署consul集群

「注意点」

• 持久化consul的数据，避免数据丢失

• 在global模式下部署Consul服务，并通过节点标签管理服务调度

• 使用两个单独的网络，一个用于内部consul的通信，另一个用于RabbitMQ与Consul服务之间的通信

consul部署清单

version: '3.6'
services:
  consul:
    image: consul:1.4.0
    hostname: "{{.Node.Hostname}}"
    networks:
      - consul
      - prod
    ports:
      - 8400:8400
      - 8500:8500
      - 8600:53
    volumes:
      - consul-data:/consul/data
    deploy:
      mode: global
      placement:
        constraints: [node.labels.consul == true]
    command: [ "agent", "-server", "-bootstrap-expect=3", "-retry-max=3", "-retry-interval=10s", "-datacenter=prod", "-join=consul", "-retry-join=consul", "-bind={{ GetInterfaceIP \"eth0\" }}", "-client=0.0.0.0", "-ui"]
networks:
  consul:
  prod:
    external: true
volumes:
  consul-data:

初始化consul集群

docker node update --label-add consul=true node1.devopsman.cn
node1.devopsman.cn
docker stack deploy -c docker-compose_consul.yaml consul
# 等待10s
root@master:~/consul_rabbitmq_docker# docker node update --label-add consul=true master.devopsman.cn
master.devopsman.cn
root@master:~/consul_rabbitmq_docker# docker node update --label-add consul=true node1.devopsman.cn
node1.devopsman.cn

root@master:~/consul_rabbitmq_docker# curl 192.168.99.134:8500/v1/status/leader
"10.0.3.5:8300"
root@master:~/consul_rabbitmq_docker# curl 192.168.99.134:8500/v1/status/peers
["10.0.3.5:8300","10.0.3.7:8300","10.0.3.9:8300"]
root@master:~/consul_rabbitmq_docker#

在浏览器打开http://192.168.99.128:8500访问consul的dashboard来验证安装是否成功。

部署RabbitMQ集群

「注意点」

• 持久化数据防止数据丢失。

• 在global模式下部署RabbitMQ服务，并通过节点标签管理服务调度

• 使用Prod网络进行内部/外部RabbitMQ通信

• 不要暴露 RABBITMQ_ERLANG_COOKIE and RABBITMQ_DEFAULT_PASS

• 主机名很重要，因为RabbitMQ使用主机名作为数据目录

RabbitMQ部署清单

version: "3.6"
services:
  rabbitmq-01:
    image: olgac/rabbitmq:3.7.8-management
    hostname: rabbitmq-01
    environment:
      - RABBITMQ_DEFAULT_USER=admin
      - RABBITMQ_DEFAULT_PASS=Passw0rd
      - RABBITMQ_ERLANG_COOKIE="MY-SECRET-KEY-123"
    networks:
      - prod
    volumes:
      - rabbitmq-01-data:/var/lib/rabbitmq
    deploy:
      mode: global
      placement:
        constraints: [node.labels.rabbitmq1 == true]rabbitmq-02:
    image: olgac/rabbitmq:3.7.8-management
    hostname: rabbitmq-02
    environment:
      - RABBITMQ_DEFAULT_USER=admin
      - RABBITMQ_DEFAULT_PASS=Passw0rd
      - RABBITMQ_ERLANG_COOKIE="MY-SECRET-KEY-123"
    networks:
      - prod
    volumes:
      - rabbitmq-02-data:/var/lib/rabbitmq
    deploy:
      mode: global
      placement:
        constraints: [node.labels.rabbitmq2 == true]rabbitmq-03:
    image: olgac/rabbitmq:3.7.8-management
    hostname: rabbitmq-03
    environment:
      - RABBITMQ_DEFAULT_USER=admin
      - RABBITMQ_DEFAULT_PASS=Passw0rd
      - RABBITMQ_ERLANG_COOKIE="MY-SECRET-KEY-123"
    networks:
      - prod
    volumes:
      - rabbitmq-03-data:/var/lib/rabbitmq
    deploy:
      mode: global
      placement:
        constraints: [node.labels.rabbitmq3 == true]networks:
  prod:
    external: true
volumes:
  rabbitmq-01-data:
  rabbitmq-02-data:
  rabbitmq-03-data:

config/enabled_plugins

[rabbitmq_management,
rabbitmq_peer_discovery_consul,
rabbitmq_federation,
rabbitmq_federation_management,
rabbitmq_shovel,
rabbitmq_shovel_management].

config/rabbitmq.conf

loopback_users.admin = false
cluster_formation.peer_discovery_backend = rabbit_peer_discovery_consul
cluster_formation.consul.host = consul
cluster_formation.node_cleanup.only_log_warning = true
cluster_formation.consul.svc_addr_auto = true
cluster_partition_handling = autoheal
vm_memory_high_watermark.relative = 0.8
disk_free_limit.absolute = 5GB

RabbitMQ集群初始化

root@master:~/consul_rabbitmq_docker# docker node update --label-add rabbitmq1=true master.devopsman.cn
master.devopsman.cn
root@master:~/consul_rabbitmq_docker# docker node update --label-add rabbitmq2=true node1.devopsman.cn
node1.devopsman.cn
root@master:~/consul_rabbitmq_docker# docker node update --label-add rabbitmq3=true node2.devopsman.cn
node2.devopsman.cn

root@master:~/consul_rabbitmq_docker# docker stack deploy -c docker-compose_rabbitmq.yml rabbitmq
Creating service rabbitmq_rabbitmq-03
Creating service rabbitmq_rabbitmq-01
Creating service rabbitmq_rabbitmq-02
root@master:~/consul_rabbitmq_docker#

配置haproxy代理端口

haproxy部署清单

version: "3.6"
services:
  haproxy:
    image: olgac/haproxy-for-rabbitmq:1.8.14-alpine
    ports:
      - 15672:15672
      - 5672:5672
      - 1936:1936
    networks:
      - prod
    deploy:
      mode: global
networks:
  prod:
    external: true

config/haproxy.cfg

global
    log 127.0.0.1   local0
    log 127.0.0.1   local1 notice
    maxconn 4096
 
defaults
    log     global
    option  tcplog
    option  dontlognull
    timeout connect 6s
    timeout client 60s
    timeout server 60s
 
listen  stats
    bind *:1936
    mode http
    stats enable
    stats hide-version
    stats realm Haproxy\ Statistics
    stats uri /listen rabbitmq
    bind   *:5672
    mode   tcp
    server rabbitmq-01 rabbitmq-01:5672 check
    server rabbitmq-02 rabbitmq-02:5672 check
    server rabbitmq-03 rabbitmq-03:5672 checklisten rabbitmq-ui
    bind   *:15672
    mode   http
    server rabbitmq-01 rabbitmq-01:15672 check
    server rabbitmq-02 rabbitmq-02:15672 check
    server rabbitmq-03 rabbitmq-03:15672 check

验证RabbitMQ集群

访问 http://192.168.99.128:15672 RabbitMQ的控制后台，密码为:(admin/Passw0rd)

最后，查看一下haproxy的后台。检查http://192.168.99.128:1936

Reference

[1]

RabbitMQ discovery consul: https://www.rabbitmq.com/cluster-formation.html#peer-discovery-consul