本文主要分为三个部分首先简单介绍一下 RocketMQ Operator 的相关知识然后结合案例详细介绍 RocketMQ Operator 提供的自定义资源及使用方法最后介绍 Operator 社区目前的情况并展望 RocketMQ Operator 下一步的发展方向。相关背景知识1. RocketMQ2012~2013 年期间阿里巴巴中间件团队自主研发并对外开源了第三代分布式消息引擎 RocketMQ其高性能、低延迟、抗堆积的特性稳定支撑了阿里巴巴 双11 万亿级数据洪峰业务其云产品 Aliware MQ 在微服务、流计算、IoT、异步解耦、数据同步等无数工况场景大放异彩。2016 年阿里巴巴向 Apache 软件基金会捐赠了 RocketMQ。次年RocketMQ 顺利从基金会毕业成为 Apache 顶级开源项目与 Apache HadoopApache Spark 一起为全球分布式、大数据领域的开发者带来福音。然而在云原生时代的今天RocketMQ 作为有状态的分布式服务系统如何在大规模集群上做到极简运维则是一个极具挑战和价值的问题。RocketMQ 支持多种部署方式以基本的双主双从架构为例如下图所示。RocketMQ 双主双从架构这里面包括了一共 7 个 RocketMQ 服务实例3 个 name server 实例2 个 master broker 实例以及 2 个 slave broker 实例。传统的部署方式需要手动或编写脚本在每个节点上进行环境和文件配置。此外随着用户业务的增加存在对集群进行无缝扩容等需求。传统方式是运维人员访问不同节点依赖操作手册和脚本按步骤进行操作来完成耗费人力且存在误操作的可能。一些公司可能会使用一些平台和工具如 Ansible 来帮助自动化运维此外越来越多的公司开始集成和使用基于 Kubernetes 的云原生生态。使用 Kubernetes 提供的 Deployment 和 StatefulSet 等原生资源可以很好地解决无状态应用的管理问题但对于数据库和 RocketMQ 这类有状态应用则存在很多局限性。例如对 RocketMQ 来说扩容不仅仅是拉起新的实例 Pod 就完成了还需要同步复制 Broker 的状态信息包括 Topic 信息和订阅关系这些元数据同时要正确配置新 Broker 的 config 参数包括 brokerName 和 NameServer IP List 等才能使得新扩容的 Broker 可用而这些仅仅靠用户编写 StatefulSet修改 size 或 replicas 然后 apply 是无法做到的。实际上 Kubernetes 开发人员也发现了这些问题因此引入了自定义资源和控制器的概念让开发人员可以直接用 Go 语言调用 Kubernetes API编写自定义资源和对应的控制器逻辑来解决复杂有状态应用的管理问题提供特定应用相关的自定义资源的这类代码组件称之为 Operator。由具备 RocketMQ 领域知识的专家编写 Operator屏蔽了应用领域的专业知识让用户只需要关心和定义希望达到的集群终态这也是 Kubernetes 声明式 API 的设计哲学。2. Kubernetes OperatorOperator 是在 Kubernetes 基础上通过扩展 Kubernetes API用来创建、配置和管理复杂的有状态应用如分布式数据库等。Operator 基于 Kubernetes 1.7 版本以来引入的自定义控制器的概念在自定义资源和控制器之上构建同时又包含了应用程序特定的领域知识。实现一个 Operator 的关键是 CRD自定义资源和 Controller控制器的设计。Operator 站在 Kubernetes 内部视角为应用的云原生化打开了新世界的大门。自定义资源可以让开发人员扩展添加新功能更新现有的功能并且可以自动执行一些管理任务这些自定义的控制器就像 Kubernetes 原生的组件一样Operator 可以直接使用 Kubernetes API 进行开发也就是说他们可以根据这些控制器编写的自定义规则来创建和更改 Pods / Services、对正在运行的应用进行扩缩容。快速开始本文使用 RocketMQ Operator 0.2.1 版本展示如何使用 RocketMQ Operator 在 Kubernetes 上快速创建部署一个 RocketMQ 服务集群。准备好 K8s 环境可以使用 docker desktop 自带的 K8s或者 minikube克隆 rocketmq-operator 仓库到你的 K8s 节点上$ git clone https://github.com/apache/rocketmq-operator.git $ cd rocketmq-operator运行脚本安装 RocketMQ Operator$ ./install-operator.sh检查下 RocketMQ Operator 是否安装成功$ kubectl get pods NAME READY STATUS RESTARTS AGE rocketmq-operator-564b5d75d-jllzk 1/1 Running 0 108s成功安装时rocketmq-operator pod 处于类似上面例子的 running 状态。应用 Broker 和 NameService 自定义资源创建 RocketMQ 集群应用 rocketmq-operator / example 中的 rocketmq_v1alpha1_rocketmq_cluster.yaml 文件快速部署一个 RocketMQ 集群。rocketmq_v1alpha1_rocketmq_cluster.yaml 文件内容如下apiVersion: rocketmq.apache.org/v1alpha1 kind: Broker metadata: # name of broker cluster name: broker spec: # size is the number of the broker cluster, each broker cluster contains a master broker and [replicaPerGroup] replica brokers. size: 1 # nameServers is the [ip:port] list of name service nameServers: # replicationMode is the broker replica sync mode, can be ASYNC or SYNC replicationMode: ASYNC # replicaPerGroup is the number of each broker cluster replicaPerGroup: 1 # brokerImage is the customized docker image repo of the RocketMQ broker brokerImage: apacherocketmq/rocketmq-broker:4.5.0-alpine # imagePullPolicy is the image pull policy imagePullPolicy: Always # resources describes the compute resource requirements and limits resources: requests: memory: 2048Mi cpu: 250m limits: memory: 12288Mi cpu: 500m # allowRestart defines whether allow pod restart allowRestart: true # storageMode can be EmptyDir, HostPath, StorageClass storageMode: EmptyDir # hostPath is the local path to store data hostPath: /data/rocketmq/broker # scalePodName is broker-[broker group number]-master-0 scalePodName: broker-0-master-0 # volumeClaimTemplates defines the storageClass volumeClaimTemplates: - metadata: name: broker-storage spec: accessModes: - ReadWriteOnce storageClassName: rocketmq-storage resources: requests: storage: 8Gi --- apiVersion: rocketmq.apache.org/v1alpha1 kind: NameService metadata: name: name-service spec: # size is the the name service instance number of the name service cluster size: 1 # nameServiceImage is the customized docker image repo of the RocketMQ name service nameServiceImage: apacherocketmq/rocketmq-nameserver:4.5.0-alpine # imagePullPolicy is the image pull policy imagePullPolicy: Always # hostNetwork can be true or false hostNetwork: true # Set DNS policy for the pod. # Defaults to ClusterFirst. # Valid values are ClusterFirstWithHostNet, ClusterFirst, Default or None. # DNS parameters given in DNSConfig will be merged with the policy selected with DNSPolicy. # To have DNS options set along with hostNetwork, you have to specify DNS policy # explicitly to ClusterFirstWithHostNet. dnsPolicy: ClusterFirstWithHostNet # resources describes the compute resource requirements and limits resources: requests: memory: 512Mi cpu: 250m limits: memory: 1024Mi cpu: 500m # storageMode can be EmptyDir, HostPath, StorageClass storageMode: EmptyDir # hostPath is the local path to store data hostPath: /data/rocketmq/nameserver # volumeClaimTemplates defines the storageClass volumeClaimTemplates: - metadata: name: namesrv-storage spec: accessModes: - ReadWriteOnce storageClassName: rocketmq-storage resources: requests: storage: 1Gi注意到这个例子中 storageMode: EmptyDir表示存储使用的是 EmptyDir数据会随着 Pod 的删除而抹去因此该方式仅供开发测试时使用。一般使用 HostPath 或 StorageClass 来对数据进行持久化存储。使用 HostPath 时需要配置 hostPath声明宿主机上挂载的目录。使用 storageClass 时需要配置 volumeClaimTemplates声明 PVC 模版。具体可参考 RocketMQ Operator 文档。应用上面的 yaml 文件输入命令$ kubectl apply -f example/rocketmq_v1alpha1_rocketmq_cluster.yaml broker.rocketmq.apache.org/broker created nameservice.rocketmq.apache.org/name-service created查看集群 Pod 状态$ kubectl get pods -owide NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES broker-0-master-0 1/1 Running 0 27s 10.1.2.27 docker-desktop none none broker-0-replica-1-0 1/1 Running 0 27s 10.1.2.28 docker-desktop none none name-service-0 1/1 Running 0 27s 192.168.65.3 docker-desktop none none rocketmq-operator-76b4b9f4db-x52mz 1/1 Running 0 3h25m 10.1.2.17 docker-desktop none none使用默认的 rocketmq_v1alpha1_rocketmq_cluster.yaml 文件配置我们看到集群中拉起了 1 个 name server 服务name-service-0和 2 个 broker 服务1 主 1 从。好啦到这里你已经成功通过 Operator 提供的自定义资源部署了一个 RocketMQ 服务集群。访问这个 RocketMQ 集群中的 Pod 来验证集群是否能正常工作使用 RocketMQ 的 tools.sh 脚本运行 Producer example$ kubectl exec -it broker-0-master-0 bash bash-4.4# sh ./tools.sh org.apache.rocketmq.example.quickstart.Producer OpenJDK 64-Bit Server VM warning: ignoring option PermSize128m; support was removed in 8.0 OpenJDK 64-Bit Server VM warning: ignoring option MaxPermSize128m; support was removed in 8.0 06:56:29.145 [main] DEBUG i.n.u.i.l.InternalLoggerFactory - Using SLF4J as the default logging framework SendResult [sendStatusSEND_OK, msgId0A0102CF007778308DB1206383920000, offsetMsgId0A0102CF00002A9F0000000000000000, messageQueueMessageQueue [topicTopicTest, brokerNamebroker-0, queueId0], queueOffset0] ... 06:56:51.120 [NettyClientSelector_1] INFO RocketmqRemoting - closeChannel: close the connection to remote address[10.1.2.207:10909] result: true bash-4.4#在另一个节点上运行 Consumer example$ kubectl exec -it name-service-0 bash bash-4.4# sh ./tools.sh org.apache.rocketmq.example.quickstart.Consumer OpenJDK 64-Bit Server VM warning: ignoring option PermSize128m; support was removed in 8.0 OpenJDK 64-Bit Server VM warning: ignoring option MaxPermSize128m; support was removed in 8.0 07:01:32.077 [main] DEBUG i.n.u.i.l.InternalLoggerFactory - Using SLF4J as the default logging framework Consumer Started. ConsumeMessageThread_1 Receive New Messages: [MessageExt [queueId0, storeSize273, queueOffset19845, sysFlag0, bornTimestamp1596768410268, bornHost/30.4.165.204:53450, storeTimestamp1596768410282, storeHost/100.81.180.84:10911, msgId6451B45400002A9F000014F96A0D6C65, commitLogOffset23061458676837, bodyCRC532471758, reconsumeTimes0, preparedTransactionOffset0, toString()Message{topicTopicTest, flag0, properties{MIN_OFFSET19844, TRACE_ONtrue, eagleTraceId1e04a5cc15967684102641001d0db0, MAX_OFFSET19848, MSG_REGIONDefaultRegion, CONSUME_START_TIME1596783715858, UNIQ_KEY1E04A5CC0DB0135FBAA421365A5F0000, WAITtrue, TAGSTagA, eagleRpcId9.1}, body[72, 101, 108, 108, 111, 32, 77, 101, 116, 97, 81, 32, 48], transactionIdnull}]] ConsumeMessageThread_4 Receive New Messages: [MessageExt [queueId1, storeSize273, queueOffset19637, sysFlag0, bornTimestamp1596768410296, bornHost/30.4.165.204:53450, storeTimestamp1596768410298, storeHost/100.81.180.84:10911, msgId6451B45400002A9F000014F96A0D7141, commitLogOffset23061458678081, bodyCRC1757146968, reconsumeTimes0, preparedTransactionOffset0, toString()Message{topicTopicTest, flag0, properties{MIN_OFFSET19636, TRACE_ONtrue, eagleTraceId1e04a5cc15967684102961002d0db0, MAX_OFFSET19638, MSG_REGIONDefaultRegion, CONSUME_START_TIME1596783715858, UNIQ_KEY1E04A5CC0DB0135FBAA421365AB80001, WAITtrue, TAGSTagA, eagleRpcId9.1}, body[72, 101, 108, 108, 111, 32, 77, 101, 116, 97, 81, 32, 49], transactionIdnull}]] ...删除集群清理环境清除 RocketMQ 服务集群实例$ kubectl delete -f example/rocketmq_v1alpha1_rocketmq_cluster.yaml清除 RocketMQ Operator$ ./purge-operator.sh按照 OperatorHub 官网指导安装 RocketMQ Operator在 OperatorHub.io 网页搜索 RocketMQ Operator选择 Streaming Messaging 类别点击 RocketMQ Operator进入 RocketMQ Operator 页面点击 Install 按钮按照说明安装 OLM 和 RocketMQ Operator本地安装 OLM 来使用 RocketMQ Operator本地安装和动 OLM(Operator Lifecycle Manager) console参考OLM 安装文档。本地启动 UI 界面控制台$ make run-console-local访问 http://localhost:9000 查看控制台OperatorHub搜索 RocketMQ 或点击 All Items 分类中的 Streaming Messaging找到 RocketMQ Operator 并进行安装安装完 RocketMQ Operator 后可以在 Installed Operators 中找到 RocketMQ Operator已安装的 Operators 界面RocketMQ Operator 介绍界面通过 UI 界面创建 NameService 自定义资源可以在 UI 中创建指定 Namespace 下的 NameService 和 Broker 实例并对已创建的实例进行浏览和管理。我们也可以通过命令查看当前 K8s 集群中的 Pod 状态例如$ kubectl get pods -A NAMESPACE NAME READY STATUS RESTARTS AGE docker compose-78f95d4f8c-8fr5z 1/1 Running 0 32h docker compose-api-6ffb89dc58-nv9rh 1/1 Running 0 32h kube-system coredns-5644d7b6d9-hv6r5 1/1 Running 0 32h kube-system coredns-5644d7b6d9-mkqb6 1/1 Running 0 32h kube-system etcd-docker-desktop 1/1 Running 0 32h kube-system kube-apiserver-docker-desktop 1/1 Running 0 32h kube-system kube-controller-manager-docker-desktop 1/1 Running 1 32h kube-system kube-proxy-snmxh 1/1 Running 0 32h kube-system kube-scheduler-docker-desktop 1/1 Running 1 32h kube-system storage-provisioner 1/1 Running 1 32h kube-system vpnkit-controller 1/1 Running 0 32h marketplace broker-0-master-0 1/1 Running 0 5h3m marketplace broker-0-replica-1-0 1/1 Running 0 5h3m marketplace name-service-0 1/1 Running 0 5h3m marketplace marketplace-operator-69756457d8-42chk 1/1 Running 0 32h marketplace rocketmq-operator-0.2.1-c9fffb5f-cztcl 1/1 Running 0 32h marketplace rocketmq-operator-84c7bb4ddc-7rvqr 1/1 Running 0 32h marketplace upstream-community-operators-5b79db455f-7t47w 1/1 Running 1 32h olm catalog-operator-7b788c597d-gjz55 1/1 Running 0 32h olm olm-operator-946bd977f-dhszg 1/1 Running 0 32h olm operatorhubio-catalog-fvxp9 1/1 Running 0 32h olm packageserver-789c7b448b-7ss7m 1/1 Running 0 32h olm packageserver-789c7b448b-lfxrw 1/1 Running 0 32h可以看到在 marketplace 这个 namespace 中也成功创建了对应的 name server 和 broker 实例。以上是基于 OperatorHub 和 OLM 安装使用 RocketMQ Operator 的案例我们将持续推送和维护新版本的 RocketMQ Operator 至该平台方便用户获取最新更新或选择合适的 Operator 版本。