使用glusterfs 作为 kubernetes PersistentVolume PersistentVolumeClaim 持久化仓库,高可用Rabbitmq,高可用mysql,高可用redis

glusterfs 怎么集群,网上一搜铺天盖地的

可利用这个特点做单节点高可用,因为K8S 哪怕节点宕机了 master 会在随意一台节点把挂掉的复活

当然我是在自己的环境下跑,经过网络的glusterfs,数据传输,有笔记大的性能损耗,对网络要求也特别高

小文件存储性能也不高等问题.最下面有mariadb 使用glusterfs 插入10W行数据,与本地硬盘的性能对比

这里记录一下rabbitmq 单机高可用情景,mysql,mongodb, redis 等,万变不离其宗

事先创建好了 volume,卷名为 env-dev

随便找个客户机挂载

mount -t glusterfs 192.168.91.135:/env-dev /mnt/env/dev      

预先创建需要的文件夹

mkdir -p /mnt/env/dev/rabbitmq/mnesia
      

　　

编写 glusterfs endpoint

[root@k8s-master-0 dev]# cat pv-ep.yaml 
apiVersion: v1
kind: Endpoints
metadata:
  name: glusterfs
  namespace: env-dev
subsets:
- addresses:
  - ip: 192.168.91.135
  - ip: 192.168.91.136
  ports:
  - port: 49152
    protocol: TCP

---
apiVersion: v1
kind: Service
metadata:
  name: glusterfs
  namespace: env-dev
spec:
  ports:
  - port: 49152
    protocol: TCP
    targetPort: 49152
  sessionAffinity: None
  type: ClusterIP      

编写 pv,注意这里path 是 volume名称 + 具体路径

[root@k8s-master-0 dev]# cat rabbitmq-pv.yaml 
apiVersion: v1
kind: PersistentVolume
metadata:
  name: rabbitmq-pv
  labels:
    type: glusterfs
spec:
  storageClassName: rabbitmq-dir
  capacity:
    storage: 3Gi
  accessModes:
    - ReadWriteMany
  glusterfs:
    endpoints: glusterfs
    path: "env-dev/rabbitmq/mnesia"
    readOnly: false      

编写pvc

[root@k8s-master-0 dev]# cat rabbitmq-pvc.yaml 
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: rabbitmq-pvc
  namespace: env-dev
spec:
  storageClassName: rabbitmq-dir
  accessModes:
    - ReadWriteMany
  resources:
    requests:
      storage: 3Gi      

创建endpoint pv pcv

kubectl apply -f pv-ep.yaml
kubectl apply -f rabbitmq-pv.yaml

kubectl apply -f rabbitmq-pvc.yaml      

使用方式,红字部分

apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: ha-rabbitmq
  namespace: env-dev
spec:
  replicas: 1
  selector:
    matchLabels:
      app: ha-rabbitmq
  template:
    metadata:
      labels:
        app: ha-rabbitmq
    spec:
      #hostNetwork: true
      hostname: ha-rabbitmq
      terminationGracePeriodSeconds: 60
      containers:
      - name: ha-rabbitmq
        image: 192.168.91.137:5000/rabbitmq:3.7.7-management-alpine
        securityContext:
          privileged: true
        env:
        - name: "RABBITMQ_DEFAULT_USER"
          value: "rabbit"
        - name: "RABBITMQ_DEFAULT_PASS"
          value: "rabbit"
        ports:
        - name: tcp
          containerPort: 5672
          hostPort: 5672
        - name: http
          containerPort: 15672
          hostPort: 15672
        livenessProbe:
          failureThreshold: 3
          httpGet:
            path: /
            port: 15672
            scheme: HTTP
          initialDelaySeconds: 20
          periodSeconds: 10
          successThreshold: 1
          timeoutSeconds: 1
        readinessProbe:
          failureThreshold: 3
          httpGet:
            path: /
            port: 15672
            scheme: HTTP
          initialDelaySeconds: 20
          periodSeconds: 10
          successThreshold: 1
          timeoutSeconds: 1
        volumeMounts:
        - name: date
          mountPath: /etc/localtime
        - name: workdir
          mountPath: "/var/lib/rabbitmq/mnesia"
      volumes:
      - name: date
        hostPath: 
          path: /usr/share/zoneinfo/Asia/Shanghai
      - name: workdir
        persistentVolumeClaim:
          claimName: rabbitmq-pvc

---

apiVersion: v1
kind: Service
metadata:
  name: ha-rabbitmq
  namespace: env-dev
  labels:
    app: ha-rabbitmq
spec:
  ports:
  - name: tcp
    port: 5672
    targetPort: 5672
  - name: http
    port: 15672
    targetPort: 15672      

创建rabbitmq pod以及service.

kubectl create -f ha-rabbitmq.yaml      

分配到了第一个节点,看看数据文件

在管理页面创建一个创建一个 virtual host

环境东西太多,这里就不暴力关机了,直接删除再创建

这次分配到节点0

看看刚才创建的virtual host

还健在

haproxy 代理

[root@localhost conf]# cat haproxy.cfg 
global
chroot /usr/local
daemon
nbproc 1
group nobody
user nobody
pidfile /haproxy.pid
#ulimit-n 65536
#spread-checks 5m
#stats timeout 5m
#stats maxconn 100

########默认配置############
defaults
mode tcp
retries 3              #两次连接失败就认为是服务器不可用，也可以通过后面设置
option redispatch      #当serverId对应的服务器挂掉后，强制定向到其他健康的服务器
option abortonclose    #当服务器负载很高的时候，自动结束掉当前队列处理比较久的链接
maxconn 32000          #默认的最大连接数
timeout connect 10s #连接超时
timeout client 8h #客户端超时
timeout server 8h #服务器超时
timeout check 10s    #心跳检测超时
log 127.0.0.1 local0 err #[err warning info debug]

########MariaDB配置#################
listen mariadb
bind 0.0.0.0:3306
mode tcp
balance leastconn
server mariadb1 192.168.91.141:3306 check port 3306 inter 2s rise 1 fall 2 maxconn 1000
server mariadb2 192.168.91.142:3306 check port 3306 inter 2s rise 1 fall 2 maxconn 1000
server mariadb3 192.168.91.143:3306 check port 3306 inter 2s rise 1 fall 2 maxconn 1000

#######RabbitMq配置#################
listen rabbitmq
bind 0.0.0.0:5672
mode tcp
balance leastconn
server rabbitmq1 192.168.91.141:5672 check port 5672 inter 2s rise 1 fall 2 maxconn 1000
server rabbitmq2 192.168.91.142:5672 check port 5672 inter 2s rise 1 fall 2 maxconn 1000
server rabbitmq3 192.168.91.143:5672 check port 5672 inter 2s rise 1 fall 2 maxconn 1000

#######Redis配置#################
listen redis
bind 0.0.0.0:6379
mode tcp
balance leastconn
server redis1 192.168.91.141:6379 check port 6379 inter 2s rise 1 fall 2 maxconn 1000
server redis2 192.168.91.142:6379 check port 6379 inter 2s rise 1 fall 2 maxconn 1000
server redis3 192.168.91.143:6379 check port 6379 inter 2s rise 1 fall 2 maxconn 1000      

 nginx 代理管理页面

Mariadb 使用 glusterfs 与 本地硬盘 性能测试

#使用glusterfs 两个副本的 mariadb 插入10W行数据

2 queries executed, 2 success, 0 errors, 0 warnings

查询：CREATE PROCEDURE test_insert() BEGIN DECLARE Y bigint DEFAULT 1; WHILE Y<100000 DO INSERT INTO t_insert VALUES(NULL,'11111111111...

共 0 行受到影响

执行耗时   : 0.018 sec
传送时间   : 1.366 sec
总耗时      : 1.385 sec
-----------------------------------------------------------

查询：CALL test_insert();

共 99999 行受到影响

执行耗时   : 6 min 1 sec
传送时间   : 0 sec
总耗时      : 6 min 1 sec


#使用本地硬盘 mariadb 插入10W行数据

2 queries executed, 2 success, 0 errors, 0 warnings

查询：CREATE PROCEDURE test_insert() BEGIN DECLARE Y BIGINT DEFAULT 1; WHILE Y<100000 DO INSERT INTO t_insert VALUES(NULL,'11111111111...

共 0 行受到影响

执行耗时   : 0.004 sec
传送时间   : 1.372 sec
总耗时      : 1.376 sec
-----------------------------------------------------------

查询：CALL test_insert();

共 99999 行受到影响

执行耗时   : 1 min 13 sec
传送时间   : 0.006 sec
总耗时      : 1 min 13 sec