MySQL高可用新玩法之MGR+Consul

前面的文章有提到過利用consul+mha實現mysql的高可用，以及利用consul+sentinel實現redis的高可用，具體的請查看：http://www.cnblogs.com/gomysql/p/8010552.html。本次給大家帶來mysql高可用的新玩法，利用mysql 5.7的mgr+consul實現，至於mgr是什么，有什么優勢，如何搭建這里就不說了，大家自己google，我這里就是介紹利用mgr+consul實現高可用及故障自動切換。至於consul是什么可以參考前面的文章。

環境：mgr至少需要3個節點。數據庫版本：mysql 5.7.19

我這里使用單主模式。

192.168.100.78
192.168.100.75
192.168.100.74

mgr搭建完成以后查看狀態：

[root@localhost][performance_schema]> select * from performance_schema.replication_group_members;
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST     | MEMBER_PORT | MEMBER_STATE |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| group_replication_applier | 8e4d93b8-4dd1-11e8-8306-6c92bf7e18e2 | ym_DB_16_100075 |        3306 | ONLINE       |
| group_replication_applier | 9062a0ef-4dd1-11e8-8822-6c92bf744dd6 | ym_DB_15_100074 |        3306 | ONLINE       |
| group_replication_applier | 9a7e7cd5-4dd1-11e8-b28c-6c92bf7e0d2e | ym_DB_19_100078 |        3306 | ONLINE       |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+

查看主節點是哪個（我的環境是單主，官方也是推薦使用單主）

[root@localhost][performance_schema]> select * from  performance_schema.replication_group_members where member_id =(select variable_value from performance_schema.global_status WHERE VARIABLE_NAME= 'group_replication_primary_member');
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST     | MEMBER_PORT | MEMBER_STATE |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| group_replication_applier | 8e4d93b8-4dd1-11e8-8306-6c92bf7e18e2 | ym_DB_16_100075 |        3306 | ONLINE       |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
1 row in set (0.00 sec)

可以看見我目前的主節點是100.75。到此mgr環境ok。下面到部署consul，前面的文章提到過，需要服務發現的機器都需要安裝consul客戶端，也就是3台服務器都需要安裝。其中涉及到2個檢查腳本（腳本不夠完善，比如復制延時是否進行注冊）。
主節點檢查腳本：

#!/bin/bash
port=$1
user="root"
passwod="123"

comm="/usr/local/mysql/bin/mysql -u$user -h 127.0.0.1 -P $port -p$passwod"
value=`$comm -Nse "select 1"`
primary_member=`$comm -Nse "select variable_value from performance_schema.global_status WHERE VARIABLE_NAME= 'group_replication_primary_member'"`
server_uuid=`$comm -Nse "select variable_value from performance_schema.global_variables where VARIABLE_NAME='server_uuid';"`


# 判斷mysql是否存活
if [ -z $value ]
then
   echo "mysql $port is down....."
   exit 2
fi


# 判斷節點狀態
node_state=`$comm -Nse "select MEMBER_STATE from performance_schema.replication_group_members where MEMBER_ID='$server_uuid'"`
if [ $node_state != "ONLINE" ]
then
   echo "MySQL $port state is not online...."
   exit 2
fi


# 判斷是不是主節點
if [[ $server_uuid == $primary_member ]]
then
   echo "MySQL $port  Instance is master ........"
   exit 0
else
   echo "MySQL $port  Instance is slave ........"
   exit 2
fi

從節點檢查腳本：

#!/bin/bash
port=$1
user="root"
passwod="123"

comm="/usr/local/mysql/bin/mysql -u$user -h 127.0.0.1 -P $port -p$passwod"
value=`$comm -Nse "select 1"`
primary_member=`$comm -Nse "select variable_value from performance_schema.global_status WHERE VARIABLE_NAME= 'group_replication_primary_member'"`
server_uuid=`$comm -Nse "select variable_value from performance_schema.global_variables where VARIABLE_NAME='server_uuid';"`


# 判斷mysql是否存活
if [ -z $value ]
then
   echo "mysql $port is down....."
   exit 2
fi

# 判斷節點狀態
node_state=`$comm -Nse "select MEMBER_STATE from performance_schema.replication_group_members where MEMBER_ID='$server_uuid'"`
if [ $node_state != "ONLINE" ]
then
   echo "MySQL $port state is not online...."
   exit 2
fi


# 判斷是不是主節點
if [[ $server_uuid != $primary_member ]]
then
   echo "MySQL $port  Instance is slave ........"
   exit 0
else
   node_num=`$comm -Nse "select count(*) from  performance_schema.replication_group_members"`
   # 判斷如果沒有任何從節點，主節點也注冊從角色服務。
   if [ $node_num -eq 1 ]
   then
       echo "MySQL $port  Instance is slave ........"
       exit 0
   else
       echo "MySQL $port  Instance is master ........"
       exit 2
   fi
fi

其中一台服務器的consul配置文件，有master和slave，如下：

[root@ym_DB_16_100075 conf]# cat payment-3306-mydb-ser.json 
{
  "services": [
    {
      "name": "payment-3306-mydb-ser",
      "tags": [
        "充值-3306"
      ],
      "address": "192.168.100.75",
      "port": 3306,
      "checks": [
        {
          "script": "/usr/local/consul/shell/check_mysql_mgr_master.sh 3306",
          "interval": "15s"
        }
      ]
    }
  ]
}

[root@ym_DB_16_100075 conf]# cat r-payment-3306-mydb-ser.json 
{
  "services": [
    {
      "name": "r-payment-3306-mydb-ser",
      "tags": [
        "充值-3306"
      ],
      "address": "192.168.100.75",
      "port": 3306,
      "checks": [
        {
          "script": "/usr/local/consul/shell/check_mysql_mgr_slave.sh 3306",
          "interval": "15s"
        }
      ]
    }
  ]
}

其他兩台服務器配置文件一樣，只是"address"改成對應服務器的地址就完事。啟動consul。ping其中一個域名，比如：payment-3306-mydb-ser.service.consul，那么返回的是主節點的ip，因為這個域名是寫的。如果ping r-payment-3306-mydb-ser.service.consul，那么返回的是另外兩個從節點的ip。

從上面可以看到寫的域名解析到的主節點，讀的域名解析到了2個從節點，從節點可以實現負載均衡的效果。

故障測試：

1. 把主節點停掉，查看寫的域名payment-3306-mydb-ser.service.consul會解析到哪里。

[root@localhost][(none)]> select * from  performance_schema.replication_group_members;
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST     | MEMBER_PORT | MEMBER_STATE |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| group_replication_applier | 9062a0ef-4dd1-11e8-8822-6c92bf744dd6 | ym_DB_15_100074 |        3306 | ONLINE       |
| group_replication_applier | 9a7e7cd5-4dd1-11e8-b28c-6c92bf7e0d2e | ym_DB_19_100078 |        3306 | ONLINE       |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
2 rows in set (0.00 sec)

停掉以后發現之前的100.75主節點已經被移除，已經自動選出新的節點：

[root@localhost][(none)]> select * from  performance_schema.replication_group_members where member_id =(select variable_value from performance_schema.global_status WHERE VARIABLE_NAME= 'group_replication_primary_member');
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST     | MEMBER_PORT | MEMBER_STATE |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+
| group_replication_applier | 9062a0ef-4dd1-11e8-8822-6c92bf744dd6 | ym_DB_15_100074 |        3306 | ONLINE       |
+---------------------------+--------------------------------------+-----------------+-------------+--------------+

可以看到現在的主節點已經是100.74，那么我們看看域名payment-3306-mydb-ser.service.consul解析到哪里：

可以看到沒有問題，正常。再看看讀域名：r-payment-3306-mydb-ser.service.consul

可以看到一切正常。后續檢查腳本可以判斷是否延時，如果延時就不注冊服務。