Rocketmq
rocketmq角色
RocketMQ由四部分构成:Producer、Consumer、Broker和NameServer
启动顺序:NameServer->Broker
为了消除单点故障,增加可靠性或增大吞吐量,可以在多台机器上部署多个nameserver和broker,并且为每个broker部署1个或多个slave
Topic & message queue:一个分布式消息队列中间件部署好以后,可以给很多个业务提供服务,同一个业务也有不同类型的消息要投递,这些不同类型的消息以不同的 Topic 名称来区分。所以发送和接收消息前,先创建topic,针对某个 Topic 发送和接收消息。有了 Topic 以后,还需要解决性能问题 。 如果一个Topic 要发送和接收的数据量非常大, 需要能支持增加并行处理的机器来提高处理速度,这时候一个 Topic 可以根据需求设置一个或多个 Message Queue, Message Queue 类似分区或 Partition 。Topic有了多个 Message Queue 后,消息可以并行地向各个Message Queue 发送,消费者也可以并行地从多个 Message Queue 读取消息并消费 。
rocketmq集群部署方式
- 单Master模式
只有一个 Master节点
优点:配置简单,方便部署
缺点:这种方式风险较大,一旦Broker重启或者宕机时,会导致整个服务不可用,不建议线上环境使用
- 多Master模式
一个集群无 Slave,全是 Master,例如 2 个 Master 或者 3 个 Master
优点:配置简单,单个Master 宕机或重启维护对应用无影响,在磁盘配置为RAID10 时,即使机器宕机不可恢复情况下,由与 RAID10磁盘非常可靠,消息也不会丢(异步刷盘丢失少量消息,同步刷盘一条不丢)。性能最高。
缺点:单台机器宕机期间,这台机器上未被消费的消息在机器恢复之前不可订阅,消息实时性会受到受到影响
- 多Master多Slave模式(异步复制)
每个 Master 配置一个 Slave,有多对Master-Slave, HA,采用异步复制方式,主备有短暂消息延迟,毫秒级。
优点:即使磁盘损坏,消息丢失的非常少,且消息实时性不会受影响,因为Master 宕机后,消费者仍然可以从 Slave消费,此过程对应用透明。不需要人工干预。性能同多 Master 模式几乎一样。
缺点: Master 宕机,磁盘损坏情况,会丢失少量消息。
- 多Master多Slave模式(同步双写)
每个 Master 配置一个 Slave,有多对Master-Slave, HA采用同步双写方式,主备都写成功,向应用返回成功。
优点:数据与服务都无单点, Master宕机情况下,消息无延迟,服务可用性与数据可用性都非常高
缺点:性能比异步复制模式略低,大约低 10%左右,发送单个消息的 RT会略高。目前主宕机后,备机不能自动切换为主机,后续会支持自动切换功能
rocketmq集群部署
rocketmq-operator都是不能很灵活的调整副本数,毕竟每个副本对应其唯一的配置的文件,不同的broker实例使用的配置文件都是不一样的,都存在差异,一旦副本变化了,就不能很好的和broker 的配置文件内容一 一对应起来。
接下来我们创建一个基于k8s部署单master以及多master部署rocketmq集群,并且只需要一个broker配置文件,多个broker实例会自动基于该broker配置文件模板,自动生成不同broker实例的broker配置文件,扩容或者伸缩rocketmq集群 nameserver或者broker副本数的时候不需要理会配置文件,仅仅是调整实例的副本即可。
备注:该方案不适合有slave节点的部署方式。
环境说明
- 提供持久存储动态供给功能
[root@k8s-master01 ~]# kubectl get storageclass
NAME PROVISIONER RECLAIMPOLICY VOLUMEBINDINGMODE ALLOWVOLUMEEXPANSION AGE
nfs-client fuseim.pri/ifs Delete Immediate false 8d
- 提供metallb功能
按官网提供的方式修改即可,链接:https://metallb.universe.tf/
# kubectl apply -f https://raw.githubusercontent.com/metallb/metallb/v0.13.5/config/manifests/metallb-native.yaml
# vim ippool.yaml
# cat ippool.yamlapiVersion: metallb.io/v1beta1
kind: IPAddressPool
metadata:name: first-poolnamespace: metallb-system
spec:addresses:- 192.168.10.100-192.168.100.110
# kubectl create -f ippool.yaml
# vim l2.yaml
# cat l2.yamlapiVersion: metallb.io/v1beta1
kind: L2Advertisement
metadata:name: examplenamespace: metallb-system
# kubectl create -f l2.yaml
- 提供ingress nginx controller
# wget https://raw.githubusercontent.com/kubernetes/ingress-nginx/controller-v1.4.0/deploy/static/provider/cloud/deploy.yaml
部署前修改362行,把Local修改为Cluster# kubectl apply -f deploy.yaml
# kubectl get ns
NAME STATUS AGE
ingress-nginx Active 8s
# kubectl get pods -n ingress-nginx
NAME READY STATUS RESTARTS AGE
ingress-nginx-admission-create-r9gqt 0/1 Completed 0 62s
ingress-nginx-admission-patch-r24d6 0/1 Completed 0 62s
ingress-nginx-controller-7844b9db77-mxrgn 1/1 Running 0 62s
# kubectl get svc -n ingress-nginx
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
ingress-nginx-controller LoadBalancer 10.96.21.124 192.168.10.100 80:31844/TCP,443:32294/TCP 49m
ingress-nginx-controller-admission ClusterIP 10.109.18.235 <none> 443/TCP 49m
- 容器镜像仓库
没有本地容器镜像仓库,可以使用hub.dockcer.com。
构建rocketmq镜像
rocketmq-namesrv和 rocketmq-broker共用同一个镜像,仅仅是启动命令和启动参数不一样,后期可灵活的通过调整启动命令和启动参数来实现不同的效果(比如通过挂载configMap的方式自定义rocketmq的配置文件,而不需要重建rocketmq的镜像。
# vim Dockerfile
# cat DockerfileFROM docker.io/library/openjdk:8u102-jdk AS JDKLABEL mail=admin@kubemsb.comRUN rm -vf /etc/localtime \&& ln -s /usr/share/zoneinfo/Asia/Shanghai /etc/localtime \&& echo "Asia/Shanghai" > /etc/timezone \&& export LANG=zh_CN.UTF-8RUN curl -k https://mirrors.tuna.tsinghua.edu.cn/apache/rocketmq/4.9.4/rocketmq-all-4.9.4-bin-release.zip \-o /tmp/rocketmq-all-4.9.4-bin-release.zip \&& unzip /tmp/rocketmq-all-4.9.4-bin-release.zip -d /tmp/ \&& mv /tmp/rocketmq-all-4.9.4-bin-release /opt/rocketmq \&& rm -rf /tmp/*RUN sed -ir '/-Xmx/c JAVA_OPT=${JAVA_OPT}' /opt/rocketmq/bin/runserver.sh \&& sed -ir '/-Xmx/c JAVA_OPT=${JAVA_OPT}' /opt/rocketmq/bin/runbroker.sh## 运行 MQ 应用时候可以通过环境变量设置 jvm 数值,如:JAVA_OPT="-server -Xms2g -Xmx2g -XX:MetaspaceSize=128m -XX:MaxMetaspaceSize=320m"ENV ROCKETMQ_HOME=/opt/rocketmqWORKDIR $ROCKETMQ_HOME
# docker build -t docker.io/nextgomsb/rocketmq:v1 . --no-cache
# docker images
REPOSITORY TAG IMAGE ID CREATED SIZE
nextgomsb/rocketmq v1 ed01df462eb3 31 seconds ago 677MB
# docker login
Login with your Docker ID to push and pull images from Docker Hub. If you don't have a Docker ID, head over to https://hub.docker.com to create one.
Username: nextgomsb 替换为自己仓库的用户名
Password: 替换为自己仓库的用户名
WARNING! Your password will be stored unencrypted in /root/.docker/config.json.
Configure a credential helper to remove this warning. See
https://docs.docker.com/engine/reference/commandline/login/#credentials-storeLogin Succeeded
# docker push docker.io/nextgomsb/rocketmq:v1
获取rocketmq-dashboard镜像
rocketmq-dashboard是一个可视化的rocketmq集群运维监控工具。
rocketmq部署描述文件编写
rocketmq-namesrv部署描述文件
# vim rocketmq-namesrv.yaml
# cat rocketmq-namesrv.yaml---
apiVersion: v1
kind: Namespace
metadata:name: rocketmq---
apiVersion: apps/v1
kind: StatefulSet
metadata:name: rocketmq-namesrvnamespace: rocketmq
spec:serviceName: rocketmq-namesrvreplicas: 2selector:matchLabels:app: rocketmq-namesrvtemplate:metadata:labels:app: rocketmq-namesrvspec:containers:- name: rocketmq-namesrv-containerimage: docker.io/nextgomsb/rocketmq:v1imagePullPolicy: IfNotPresentcommand:- bin/mqnamesrvenv:- name: JAVA_OPTvalue: -server -Xms2g -Xmx2g -XX:MetaspaceSize=256m -XX:MaxMetaspaceSize=512m
---
apiVersion: v1
kind: Service
metadata:name: rocketmq-namesrvnamespace: rocketmqlabels:app: rocketmq-namesrv
spec:ports:- port: 9876protocol: TCPtargetPort: 9876selector:app: rocketmq-namesrvtype: ClusterIP
# dig -t a rocketmq-namesrv.rocketmq.svc.cluster.local. @10.96.0.10
rocketmq-broker部署描述文件
# vim rocketmq-broker.yaml
# cat rocketmq-broker.yaml---
apiVersion: apps/v1
kind: StatefulSet
metadata:name: rocketmq-brokernamespace: rocketmq
spec:serviceName: rocketmq-brokerreplicas: 2selector:matchLabels:app: rocketmq-brokertemplate:metadata:labels:app: rocketmq-brokerspec:containers:- name: rocketmq-brokerimage: nextgomsb/rocketmq:v1imagePullPolicy: IfNotPresentcommand:- bin/mqbroker- --namesrvAddr=rocketmq-namesrv.rocketmq.svc.cluster.local.:9876env:- name: JAVA_OPTvalue: -server -Xms1g -Xmx1gdnsPolicy: ClusterFirstrestartPolicy: AlwaysschedulerName: default-schedulerterminationGracePeriodSeconds: 30updateStrategy:rollingUpdate:partition: 0type: RollingUpdate
rocketmq-dashboard部署描述文件
部署一个能实现运维监控rocketmq的可视化web应用。
部署rocketmq-dashboard应用时候重点关注部署文件里面的env环境变量参数JAVA_OPTS,该env环境变量(JAVA_OPTS)决定了应用是否能成功连接到 rocketmq-namesrv 服务。
# vim rocketmq-dashboard.yaml
# cat rocketmq-dashboard.yaml---
apiVersion: apps/v1
kind: Deployment
metadata:name: rocketmq-dashboardnamespace: rocketmqlabels:app: rocketmq-dashboard
spec:replicas: 1selector:matchLabels:app: rocketmq-dashboardtemplate:metadata:labels:app: rocketmq-dashboardspec:containers:- name: rocketmq-dashboardimage: apacherocketmq/rocketmq-dashboard:latestimagePullPolicy: IfNotPresentenv:- name: JAVA_OPTSvalue: -Drocketmq.namesrv.addr=rocketmq-namesrv.rocketmq.svc.cluster.local.:9876dnsPolicy: ClusterFirstrestartPolicy: AlwaysschedulerName: default-schedulersecurityContext: {}terminationGracePeriodSeconds: 30---
apiVersion: v1
kind: Service
metadata:name: rocketmq-dashboardnamespace: rocketmqlabels:app: rocketmq-dashboard
spec:ports:- port: 8080protocol: TCPtargetPort: 8080selector:app: rocketmq-dashboardtype: ClusterIP
执行部署描述文件
rocketmq-namesrv部署
# kubectl create -f rocketmq-namesrv.yaml
# kubectl get pods -n rocketmq
NAME READY STATUS RESTARTS AGE
...
rocketmq-namesrv-0 1/1 Running 0 14m
rocketmq-namesrv-1 1/1 Running 0 13m
rocketmq-broker部署
# kubectl create -f rocketmq-broker.yaml
# kubectl get pods -n rocketmq
NAME READY STATUS RESTARTS AGE
rocketmq-broker-0 1/1 Running 0 4m16s
rocketmq-broker-1 1/1 Running 0 4m15s
rocketmq-dashboard部署
# kubectl create -f rocketmq-dashboard.yaml
# kubectl get pods -n rocketmq
NAME READY STATUS RESTARTS AGE
rocketmq-dashboard-f4ccdf496-sv984 1/1 Running 0 73s
调整副本
调整rocket-namesrv副本数kubectl scale sts rocketmq-namesrv --replicas=3 -n rocketmq
# kubectl get pods -n rocketmq
NAME READY STATUS RESTARTS AGE
rocketmq-namesrv-0 1/1 Running 0 15m
rocketmq-namesrv-1 1/1 Running 0 14m
rocketmq-namesrv-2 1/1 Running 0 4s
调整rocket-broker副本数,集群节点内存要大于副本数中内存。
kubectl scale sts rocketmq-broker --replicas=3 -n rocketmq
创建ingress资源对象实现域名访问dashboard
# vim rocketmq-dashboard-ingress.yaml
# cat rocketmq-dashboard-ingress.yaml
apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:name: ingress-rocketmq-dashboard #自定义ingress名称namespace: rocketmqannotations:ingressclass.kubernetes.io/is-default-class: "true"kubernetes.io/ingress.class: nginx
spec:rules:- host: rocketmq-dashboard.kubemsb.com # 自定义域名http:paths:- pathType: Prefixpath: "/"backend:service:name: rocketmq-dashboard # 对应上面创建的service名称port:number: 8080
# kubectl create -f rocketmq-dashboard-ingress.yaml
# kubectl get ingress -n rocketmq
NAME CLASS HOSTS ADDRESS PORTS AGE
ingress-rocketmq-dashboard <none> rocketmq-dashboard.kubemsb.com 80 31s