本文介绍在kubernetes环境中部署mysql主从集群，数据持久化采用nfs。

一、环境介绍
Mysql版本：5.7

Mysql master节点：
主机名：vm1
IP地址：192.168.115.5/24

Mysql slave节点：
主机名：vm2
IP地址：192.168.115.6/24

NFS节点：
主机名：vm2
IP地址：192.168.115.6/24
共享目录：/home/mysql_master、/home/mysql_slave

二、准备mysql主从的镜像环境
dockerfile、docker-entrypoint.sh 文件下载地址如下
https://github.com/docker-library/mysql/tree/master/5.7
由于我们要配置mysql主从，所以需要对dockerfile、docker-entrypoint.sh 文件做一点的修改，主要是在mysql主从配置部分。

准备master的镜像
将Dockerfile, docker-entrypoint.sh复制一份用于build master镜像文件。
在Dockerfile中添加如下内容，将mysql master的server-id设置为1

RUN sed -i '/\[mysqld\]/a server-id=1\nlog-bin' /etc/mysql/mysql.conf.d/mysqld.cnf

在docker-entrypoint.sh中添加如下内容，创建一个复制用户并赋权限，刷新系统权限表

echo "CREATE USER '$MYSQL_REPLICATION_USER'@'%' IDENTIFIED BY '$MYSQL_REPLICATION_PASSWORD' ;" | "${mysql[@]}"
echo "GRANT REPLICATION SLAVE ON *.* TO '$MYSQL_REPLICATION_USER'@'%' IDENTIFIED BY '$MYSQL_REPLICATION_PASSWORD' ;" | "${mysql[@]}"
echo 'FLUSH PRIVILEGES ;' | "${mysql[@]}"

准备slave的镜像
将Dockerfile, docker-entrypoint.sh复制一份用于build slave镜像文件。
在Dockerfile中添加如下内容，将mysql slave的server-id设置为一个随机数

RUN RAND="$(date +%s | rev | cut -c 1-2)$(echo ${RANDOM})" && sed -i '/\[mysqld\]/a server-id='$RAND'\nlog-bin' /etc/mysql/mysql.conf.d/mysqld.cnf

在docker-entrypoint.sh中添加如下内容，配置连接master主机的host、user、password等参数，并启动复制进程。

echo "STOP SLAVE;" | "${mysql[@]}" echo "CHANGE MASTER TO master_host='$MYSQL_MASTER_SERVICE_HOST', master_user='$MYSQL_REPLICATION_USER', master_password='$MYSQL_REPLICATION_PASSWORD' ;" | "${mysql[@]}" echo "START SLAVE;" | "${mysql[@]}"

三、开始使用修改好的dockerfile创建mysql master和slave镜像

# cd /root/kubernetes/lnmp/mysql/Dockerfiles/Master
# docker build -t registry.fjhb.cn/mysql-master:0.1 .
# cd /root/kubernetes/lnmp/mysql/Dockerfiles/Slave
# docker build -t registry.fjhb.cn/mysql-slave:0.1 .
# docker push registry.fjhb.cn/mysql-master:0.1
# docker push registry.fjhb.cn/mysql-slave:0.1

四、创建pv和pvc，用于mysql主从存储持久化数据
在nfs服务器上把目录创建出来

# cd /home/
# mkdir mysql_master mysql_slave

通过yaml文件创建出两组pv和pvc

# cd /root/kubernetes/lnmp/mysql/Storage
# cat nfs-pv-master.yaml
apiVersion: v1
kind: PersistentVolume
metadata:name: pv-nfs-mysql-master
spec:capacity:storage: 5Gi accessModes:- ReadWriteOncenfs: path: /home/mysql_masterserver: 192.168.115.6persistentVolumeReclaimPolicy: Recycle # cat nfs-pvc-master.yaml
kind: PersistentVolumeClaim
apiVersion: v1
metadata:name: pv-nfs-mysql-master
spec:accessModes:- ReadWriteOnceresources:requests:storage: 5Gi
# cat nfs-pv-slave.yaml
apiVersion: v1
kind: PersistentVolume
metadata:name: pv-nfs-mysql-slave
spec:capacity:storage: 6Gi accessModes:- ReadWriteOncenfs: path: /home/mysql_slaveserver: 192.168.115.6persistentVolumeReclaimPolicy: Recycle # cat nfs-pvc-slave.yaml
kind: PersistentVolumeClaim
apiVersion: v1
metadata:name: pv-nfs-mysql-slave
spec:accessModes:- ReadWriteOnceresources:requests:storage: 6Gi# kubectl create -f nfs-pv-master.yaml
# kubectl create -f nfs-pvc-master.yaml
# kubectl create -f nfs-pv-slave.yaml
# kubectl create -f nfs-pvc-slave.yaml 

五、根据yaml文件创建mysql master ReplicationController和services

# cat mysql-master-rc.yaml
apiVersion: v1
kind: ReplicationController
metadata:name: mysql-masterlabels:name: mysql-master
spec:replicas: 1selector:name: mysql-mastertemplate:metadata:labels:name: mysql-masterspec:containers:- name: mysql-masterimage: registry.fjhb.cn/mysql-master:0.1volumeMounts:- mountPath: /var/lib/mysqlname: mysql-master-dataports:- containerPort: 3306env:- name: MYSQL_ROOT_PASSWORDvalue: "12345678"- name: MYSQL_REPLICATION_USERvalue: "repl"- name: MYSQL_REPLICAITON_PASSWORDvalue: "12345678"volumes:- name: mysql-master-datapersistentVolumeClaim:claimName: pv-nfs-mysql-master# cat mysql-master-svc.yaml
apiVersion: v1
kind: Service
metadata:name: mysql-masterlabels:name: mysql-master
spec:type: NodePortports:- port: 3306targetPort: 3306name: httpnodePort: 30066selector:name: mysql-master# kubectl create -f mysql-master-rc.yaml
# kubectl create -f mysql-master-svc.yaml

使用mysql客户端连接master测试

六、根据yaml文件创建mysql slave ReplicationController和services

# cat mysql-slave-rc.yaml
apiVersion: v1
kind: ReplicationController
metadata:name: mysql-slavelabels:name: mysql-slave
spec:replicas: 1selector:name: mysql-slavetemplate:metadata:labels:name: mysql-slavespec:containers:- name: mysql-slaveimage: registry.fjhb.cn/mysql-slave:0.1volumeMounts:- mountPath: /var/lib/mysqlname: mysql-slave-dataports:- containerPort: 3306env:- name: MYSQL_ROOT_PASSWORDvalue: "12345678"- name: MYSQL_REPLICATION_USERvalue: "repl"- name: MYSQL_REPLICAITON_PASSWORDvalue: "12345678"volumes:- name: mysql-slave-datapersistentVolumeClaim:claimName: pv-nfs-mysql-slave# cat mysql-slave-svc.yaml
apiVersion: v1
kind: Service
metadata:name: mysql-slavelabels:name: mysql-slave
spec:type: NodePortports:- port: 3306targetPort: 3306name: httpnodePort: 30067selector:name: mysql-slave# kubectl create -f mysql-slave-rc.yaml
# kubectl create -f mysql-slave-svc.yaml 

七、测试与排错
通过mysql命令行连接slave查看复制的状态，发现状态是connecting

使用slave 容器连接master,发现使用root账号可以正常连接，而使用repl账号无法正常连接，提示access deny

修改master上的repl用户密码

mysql> alter user repl@'%' identified by "12345678";
Query OK, 0 rows affected (0.00 sec)
mysql> flush privileges;
Query OK, 0 rows affected (0.00 sec)

完成上述修改后在slave上进行验证一下，repl账号可以正常连接

在slave上进行change master操作，因为创建镜像的时候并未配置GTID参数，所以这里不能使用MASTER_AUTO_POSITION=1参数，如果要开启，需要去修改Dockerfile，重新生成镜像。

mysql> stop slave;
mysql> set global sql_slave_skip_counter=1;
CHANGE MASTER TO \
MASTER_HOST='mysql-master', \
MASTER_USER='repl', \
MASTER_PASSWORD='12345678';
mysql> start slave;

参考文档：
https://www.jianshu.com/p/509b65e9a4f5