基于CentOS7.9搭建MySQL高可用集群【MGR单主】

高可用

MGR单主模式是MySQL官方推出的高可用解决方案，它通过Paxos协议实现了真正的多副本强一致性，在保证数据安全的同时提供自动故障转移能力。

注：缺点是故障恢复节点时不能增量恢复，要删除原有数据，从0开始同步数据，小数据量场景问题不大

服务器信息

节点	IP地址	主机名	角色	Server ID
Node1	192.168.200.100	mysql01	主节点	1
Node2	192.168.200.101	mysql02	从节点	2
Node3	192.168.200.102	mysql03	从节点	3

软件版本

CentOS 7.9
MySQL 8.0.30
MGR 单主模式

基础环境准备（所有节点执行）

环境初始化

bash 复制代码

vi init.sh

bash 复制代码

#!/bin/bash

# 定义节点信息
NODES=("192.168.200.100 mysql01" "192.168.200.101 mysql02" "192.168.200.102 mysql03")

# 定义当前节点的密码（默认集群统一密码）
HOST_PASS="000000"

# 时间同步的目标节点
TIME_SERVER= mysql01

# 时间同步的地址段
TIME_SERVER_IP= 192.160.200.0/24

# 欢迎界面
cat > /etc/motd <<EOF
 ################################
 #   Welcome  to  mysqlcluster  #
 ################################
EOF
# 配置仓库，使用本地离线源
mount /dev/sr0 /mnt/
mv /etc/yum.repos.d/* /media/
cat > /etc/yum.repos.d/centos.repo << eof
[c]
name=c
baseurl=file:///mnt
gpgcheck=0
enabled=1
eof

# 优化ssh连接
sed -i -e 's/#UseDNS yes/UseDNS no/g' -e 's/GSSAPIAuthentication yes/GSSAPIAuthentication no/g' /etc/ssh/sshd_config
systemctl reload sshd

# 修改主机名
for node in "${NODES[@]}"; do
  ip=$(echo "$node" | awk '{print $1}')
  hostname=$(echo "$node" | awk '{print $2}')

  # 获取当前节点的主机名和 IP
  current_ip=$(hostname -I | awk '{print $1}')
  current_hostname=$(hostname)

  # 检查当前节点与要修改的节点信息是否匹配
  if [[ "$current_ip" == "$ip" && "$current_hostname" != "$hostname" ]]; then
    echo "Updating hostname to $hostname on $current_ip..."
    hostnamectl set-hostname "$hostname"

    if [ $? -eq 0 ]; then
      echo "Hostname updated successfully."
    else
      echo "Failed to update hostname."
    fi

    break
  fi
done

# 遍历节点信息并添加到 hosts 文件
for node in "${NODES[@]}"; do
  ip=$(echo "$node" | awk '{print $1}')
  hostname=$(echo "$node" | awk '{print $2}')

  # 检查 hosts 文件中是否已存在相应的解析
  if grep -q "$ip $hostname" /etc/hosts; then
    echo "Host entry for $hostname already exists in /etc/hosts."
  else
    # 添加节点的解析条目到 hosts 文件
    sudo sh -c "echo '$ip $hostname' >> /etc/hosts"
    echo "Added host entry for $hostname in /etc/hosts."
  fi
done

if [[ ! -s ~/.ssh/id_rsa.pub ]]; then
    ssh-keygen -t rsa -N '' -f ~/.ssh/id_rsa -q -b 2048
fi

# 检查并安装 expect 工具
if ! which expect &> /dev/null; then
    echo "expect 工具未安装，正在安装 expect..."
    sudo yum install -y expect
fi

# 遍历所有节点
for node in "${NODES[@]}"; do
    ip=$(echo "$node" | awk '{print $1}')
    hostname=$(echo "$node" | awk '{print $2}')

    expect -c "
        set timeout -1
        spawn ssh-copy-id -i /root/.ssh/id_rsa.pub $hostname
        expect {
            \"*password:*\" { send -- \"$HOST_PASS\r\"; exp_continue }
            \"*(yes/no)*\" { send -- \"yes\r\"; exp_continue }
            eof { exit 1 }
        }
    "
done

# 时间同步
if [[ $name == $TIME_SERVER ]]; then
    # 配置当前节点为时间同步源
    sed -i '3,6s/^/#/g' /etc/chrony.conf
    sed -i "7s/^/server $TIME_SERVER iburst/g" /etc/chrony.conf
    echo "allow $TIME_SERVER_IP" >> /etc/chrony.conf
    echo "local stratum 10" >> /etc/chrony.conf
else
    # 配置当前节点同步到目标节点
    sed -i '3,6s/^/#/g' /etc/chrony.conf
    sed -i "7s/^/server $TIME_SERVER iburst/g" /etc/chrony.conf
fi

# 重启并启用 chrony 服务
systemctl restart chronyd
systemctl enable chronyd

echo "###############################################################"
echo "#################      集群初始化成功     ######################"
echo "###############################################################"

关闭防火墙和selinux（生产环境则放通端口）

bash 复制代码

systemctl stop firewalld
setenforce 0

安装MySQL（所有节点执行）

解压软件包

bash 复制代码

tar -xf mysql-8.0.30-1.el7.x86_64.rpm-bundle.tar -C /opt/

复制给其他节点

bash 复制代码

scp -r /opt/* mysql02:/opt/
scp -r /opt/* mysql03:/opt/

安装并初始化

bash 复制代码

#卸载节点自带的mariadb
rpm -e --nodeps mariadb-libs-*

#安装mysql
 yum install -y /opt/*

#初始化mysql
mysqld --initialize --console

# 授权目录权限
chown -R mysql:mysql /var/lib/mysql/

# 启动并配置开机自启
systemctl enable --now mysqld

修改登录密码

bash 复制代码

# 获取初始密码
INIT_PASS=$(grep 'temporary password' /var/log/mysqld.log | awk '{print $NF}')
echo "初始密码: $INIT_PASS"

# 修改root密码
mysqladmin -uroot -p"$INIT_PASS" password '000000'

设置远程访问

bash 复制代码

mysql -uroot -p000000

use mysql;

update user set Host='%' where User='root';

flush privileges;

修改配置文件

节点1配置 (mysql01 - 主节点)

bash 复制代码

cat > /etc/my.cnf << 'EOF'
[mysqld]
# 基础配置
datadir=/var/lib/mysql
socket=/var/lib/mysql/mysql.sock
log-error=/var/log/mysqld.log
pid-file=/var/run/mysqld/mysqld.pid

# 网络配置
bind-address=0.0.0.0
port=3306

# 字符集
character-set-server=utf8mb4
collation-server=utf8mb4_unicode_ci

# ========== 复制必需配置 ==========
server_id=1
gtid_mode=ON
enforce_gtid_consistency=ON
log_bin=mysql-bin
binlog_format=ROW
binlog_checksum=NONE
log_slave_updates=ON

# ========== MGR配置 ==========
disabled_storage_engines="MyISAM,BLACKHOLE,FEDERATED,ARCHIVE,MEMORY"
plugin_load_add='group_replication.so'
transaction_write_set_extraction=XXHASH64
group_replication_group_name="aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa"
group_replication_start_on_boot=OFF
group_replication_local_address="192.168.200.100:33061"
group_replication_group_seeds="192.168.200.100:33061,192.168.200.101:33061,192.168.200.102:33061"
group_replication_bootstrap_group=OFF
group_replication_single_primary_mode=ON
group_replication_enforce_update_everywhere_checks=OFF
group_replication_consistency=EVENTUAL

# 性能优化
master_info_repository=TABLE
relay_log_info_repository=TABLE
relay_log_recovery=ON
slave_preserve_commit_order=ON
binlog_transaction_dependency_tracking=WRITESET

# 二进制日志
expire_logs_days=7
sync_binlog=1
innodb_flush_log_at_trx_commit=1

# InnoDB优化
innodb_buffer_pool_size=1G
innodb_log_file_size=256M
innodb_flush_method=O_DIRECT

[mysql]
default-character-set=utf8mb4
prompt=\\u@\\h [\\d]>\\_

[client]
default-character-set=utf8mb4
EOF

节点2配置 (mysql02 - 从节点)

bash 复制代码

scp /etc/my.cnf mysql02:/etc/my.cnf

bash 复制代码

# 仅修改以下参数，其他与节点1相同
server_id=2
group_replication_local_address="192.168.200.101:33061"

节点3配置 (mysql03 - 从节点)

bash 复制代码

scp /etc/my.cnf mysql03:/etc/my.cnf

bash 复制代码

# 仅修改以下参数，其他与节点1相同
server_id=3
group_replication_local_address="192.168.200.102:33061"

重启所有集群

bash 复制代码

systemctl restart mysqld

配置MGR集群

所有节点基础配置

bash 复制代码

mysql -uroot -p000000

sql 复制代码

-- 创建复制用户（使用mysql_native_password避免认证问题）
SET SQL_LOG_BIN=0;
CREATE USER 'repl'@'%' IDENTIFIED WITH mysql_native_password BY 'ReplPassword123!';
GRANT REPLICATION SLAVE ON *.* TO 'repl'@'%';
GRANT BACKUP_ADMIN ON *.* TO 'repl'@'%';
GRANT GROUP_REPLICATION_STREAM ON *.* TO 'repl'@'%';
SET SQL_LOG_BIN=1;
FLUSH PRIVILEGES;

验证插件

bash 复制代码

SELECT PLUGIN_NAME, PLUGIN_STATUS FROM INFORMATION_SCHEMA.PLUGINS WHERE PLUGIN_NAME = 'group_replication';

bash 复制代码

root@localhost [(none)]> SELECT PLUGIN_NAME, PLUGIN_STATUS
    -> FROM INFORMATION_SCHEMA.PLUGINS
    -> WHERE PLUGIN_NAME = 'group_replication';
+-------------------+---------------+
| PLUGIN_NAME       | PLUGIN_STATUS |
+-------------------+---------------+
| group_replication | ACTIVE        |
+-------------------+---------------+
1 row in set (0.00 sec)

节点1初始化集群（主节点）

sql 复制代码

-- 仅在mysql01上执行
SET GLOBAL group_replication_bootstrap_group=ON;
START GROUP_REPLICATION;
SET GLOBAL group_replication_bootstrap_group=OFF;

-- 验证
SELECT 
    MEMBER_HOST,
    MEMBER_STATE,
    MEMBER_ROLE,
    IF(MEMBER_STATE='ONLINE', '✅ 成功', '❌ 失败') as 状态
FROM performance_schema.replication_group_members;

bash 复制代码

root@localhost [(none)]> SELECT
    ->     MEMBER_HOST,
    ->     MEMBER_STATE,
    ->     MEMBER_ROLE,
    ->     IF(MEMBER_STATE='ONLINE', '\U+2705 成功', '\U+274C 失败') as 状态
    -> FROM performance_schema.replication_group_members;
+-------------+--------------+-------------+------------+
| MEMBER_HOST | MEMBER_STATE | MEMBER_ROLE | 状态       |
+-------------+--------------+-------------+------------+
| mysql01     | ONLINE       | PRIMARY     | ✅ 成功     |
+-------------+--------------+-------------+------------+
1 row in set (0.00 sec)

节点2加入集群

sql 复制代码

-- 在 mysql02 上执行
mysql -uroot -p000000

RESET MASTER;
START GROUP_REPLICATION USER='repl', PASSWORD='ReplPassword123!';

节点3加入集群

sql 复制代码

-- 在 mysql03 上执行
mysql -uroot -p000000

RESET MASTER;
START GROUP_REPLICATION USER='repl', PASSWORD='ReplPassword123!';

验证集群状态

查看集群成员

sql 复制代码

-- 在任一节点执行
SELECT * FROM performance_schema.replication_group_members;

bash 复制代码

root@localhost [(none)]> SELECT * FROM performance_schema.replication_group_members;
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+----------------------------+
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION | MEMBER_COMMUNICATION_STACK |
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+----------------------------+
| group_replication_applier | 3afb0ef2-fbf5-11f0-bb0f-000c29eb9544 | mysql01     |        3306 | ONLINE       | PRIMARY     | 8.0.30         | XCom                       |
| group_replication_applier | 44cf7653-fbf5-11f0-8efe-000c29e0f099 | mysql02     |        3306 | ONLINE       | SECONDARY   | 8.0.30         | XCom                       |
| group_replication_applier | 45969291-fbf5-11f0-8e9d-000c295639c3 | mysql03     |        3306 | ONLINE       | SECONDARY   | 8.0.30         | XCom                       |
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+----------------------------+
3 rows in set (0.00 sec)

root@localhost [(none)]>

检查读写状态

sql 复制代码

SELECT @@hostname, @@read_only, @@super_read_only;

bash 复制代码

-- 在主节点
root@localhost [(none)]> SELECT @@hostname, @@read_only, @@super_read_only;
+------------+-------------+-------------------+
| @@hostname | @@read_only | @@super_read_only |
+------------+-------------+-------------------+
| mysql01    |           0 |                 0 |
+------------+-------------+-------------------+

bash 复制代码

-- 在从节点
root@localhost [(none)]> SELECT @@hostname, @@read_only, @@super_read_only;
+------------+-------------+-------------------+
| @@hostname | @@read_only | @@super_read_only |
+------------+-------------+-------------------+
| mysql02    |           1 |                 1 |
+------------+-------------+-------------------+
1 row in set (0.01 sec)

测试数据同步

sql 复制代码

-- 在主节点执行
CREATE DATABASE mgr_test;
USE mgr_test;
CREATE TABLE cluster_test (
    id INT AUTO_INCREMENT PRIMARY KEY,
    node_name VARCHAR(50),
    created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

INSERT INTO cluster_test (node_name) VALUES ('主节点写入测试');

bash 复制代码

-- 在从节点验证
SELECT * FROM mgr_test.cluster_test;

bash 复制代码

root@localhost [(none)]> SELECT * FROM mgr_test.cluster_test;
+----+-----------------------+---------------------+
| id | node_name             | created_at          |
+----+-----------------------+---------------------+
|  1 | 主节点写入测试        | 2026-01-28 00:41:32 |
+----+-----------------------+---------------------+
1 row in set (0.00 sec)

模拟主节点故障

停止主节点MySQL

bash 复制代码

[root@mysql01 ~]# systemctl stop mysqld

bash 复制代码

# 在其他节点检查新主节点选举
mysql -uroot -p000000 -e "
SELECT 
    MEMBER_HOST,
    MEMBER_STATE,
    MEMBER_ROLE,
    CASE 
        WHEN MEMBER_ROLE = 'PRIMARY' THEN '🟢 新主节点'
        ELSE '🔵 从节点'
    END as 状态
FROM performance_schema.replication_group_members;"

bash 复制代码

root@localhost [(none)]> SELECT
    ->     MEMBER_HOST,
    ->     MEMBER_STATE,
    ->     MEMBER_ROLE,
    ->     CASE
    ->         WHEN MEMBER_ROLE = 'PRIMARY' THEN '\U+1F7E2 新主节点'
    ->         ELSE '\U+1F535 从节点'
    ->     END as 状态
    -> FROM performance_schema.replication_group_members;
+-------------+--------------+-------------+-------------------+
| MEMBER_HOST | MEMBER_STATE | MEMBER_ROLE | 状态              |
+-------------+--------------+-------------+-------------------+
| mysql02     | ONLINE       | PRIMARY     | 🟢 新主节点         |
| mysql03     | ONLINE       | SECONDARY   | 🔵 从节点           |
+-------------+--------------+-------------+-------------------+
2 rows in set (0.00 sec)

测试新主节点mysql02写入

bash 复制代码

mysql -uroot -p000000
USE mgr_test;
INSERT INTO cluster_test (node_name) VALUES ('故障转移后写入测试');
SELECT * FROM cluster_test;

mysql03节点查看

bash 复制代码

root@localhost [(none)]> SELECT * FROM cluster_test;
ERROR 1046 (3D000): No database selected
root@localhost [(none)]> use mgr_test;
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A

Database changed
root@localhost [mgr_test]> SELECT * FROM cluster_test;
+----+-----------------------------+---------------------+
| id | node_name                   | created_at          |
+----+-----------------------------+---------------------+
|  1 | 主节点写入测试              | 2026-01-28 00:41:32 |
|  2 | 故障转移后写入测试          | 2026-01-28 00:50:23 |
+----+-----------------------------+---------------------+
2 rows in set (0.00 sec)

root@localhost [mgr_test]>

恢复原主节点（此时则需要先删除本地原有的数据再同步，不然会冲突）
MGR 增量恢复（Incremental Recovery）的典型问题。节点有本地数据，但 GTID 位置不同，导致恢复时出现冲突

bash 复制代码

# 启动原主节点
systemctl start mysqld

mysql -uroot -p000000

# 删除原有数据
drop database mgr_test;

# 重新加入集群
RESET MASTER;
START GROUP_REPLICATION USER='repl', PASSWORD='ReplPassword123!';

查询恢复数据

bash 复制代码

root@localhost [mgr_test]> select * from cluster_test;
+----+-----------------------------+---------------------+
| id | node_name                   | created_at          |
+----+-----------------------------+---------------------+
|  1 | 主节点写入测试              | 2026-01-28 00:41:32 |
|  2 | 故障转移后写入测试          | 2026-01-28 00:50:23 |
+----+-----------------------------+---------------------+