[MariaDB]MHA高可用部署-实验

目录

• 一、简介
  • 1.1MHA角色
• 二、MHA的工具
• 三、MHA部署过程
  • 3.1.1 配置
  • 3.1.2 环境规划
  • 3.1.3 配置一主多从
  • 3.2 MHA配置
  • 3.2.1 master权限授予
  • 3.2.2 ssh互信
  • 3.2.3 安装mha包
  • 3.2.4 MHA管理节点配置
  • 3.2.5 MHA节点检测
  • 3.2.6 MHA启动
  • 3.2.7 MHA模拟故障
  • 3.2.8 修复已损坏的节点加入MHA中
• MHA 问题集锦
  • 错误一
  • 错误二
  • 错误三

一、简介

MHA的逻辑是，为了保证其MySQL的高可用，会有一个StandBy状态的master.在mysql故障切换的过程中，MHA 能做到在 0~30 秒内自动完成数据库的故障切换操作，并且在进行故障切换的过程中，MHA 能最大程度的保证数据的一致性，以达到相对意义上的高可用。

1.1MHA角色

如下图，整个 MHA 架构分为

• MHA Manager 节点
• MHA Node 节点
  
  其中 MHA Manager 节点是单点部署，MHA Node 节点是部署在每个需要监控的 MySQL 集群节点上的。MHA Manager 会定时探测集群中的 Master 节点，当 Master 出现故障时，它可以自动将最新数据的 Standby Master 或 Slave 提升为新的 Master，然后将其他的 Slave 重新指向新的 Master。

二、MHA的工具

Manager节点：

• masterha_check_ssh：MHA 依赖的 ssh 环境监测工具；
• masterha_check_repl：MYSQL 复制环境检测工具；
• masterga_manager：MHA 服务主程序；
• masterha_check_status：MHA 运行状态探测工具；
• masterha_master_monitor：MYSQL master 节点可用性监测工具；
• masterha_master_swith:master：节点切换工具；
• masterha_conf_host：添加或删除配置的节点；
• masterha_stop：关闭 MHA 服务的工具。

Node节点：（这些工具通常由MHA Manager的脚本触发，无需人为操作）

• save_binary_logs：保存和复制 master 的二进制日志；
• apply_diff_relay_logs：识别差异的中继日志事件并应用于其他 slave；
• purge_relay_logs：清除中继日志（不会阻塞 SQL 线程）；

自定义扩展：

• secondary_check_script：通过多条网络路由检测master的可用性；
• master_ip_failover_script：更新application使用的masterip；
• report_script：发送报告;
• init_conf_load_script：加载初始配置参数；
• master_ip_online_change_script;更新master节点ip地址。

三、MHA部署过程

3.1.1 配置

MHA 对 MYSQL 复制环境有特殊要求，例如各节点都要开启二进制日志及中继日志，各从节点必须显示启用其read-only属性，并关闭relay_log_purge功能等，这里对配置做事先说明。

3.1.2 环境规划

机器名	IP	角色	备注
manager	172.30.200.100	manager控制器	用于管理和故障切换
master	172.30.200.101	数据库主服务器	开启binlog,relay-log。关闭relay_log_purge
slave1	172.30.200.102	数据库从服务器	开启binlog,relay-log。关闭relay_log_purge
slave2	172.30.200.103	数据库从服务器	开启binlog,relay-log。关闭relay_log_purge

在各个节点的/etc/hosts文件配置内容添加如下:

[root@localhost ~]# cat /etc/hosts
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
172.30.200.100  arpmgr
172.30.200.101  arpmaster
172.30.200.102  arpslave1
172.30.200.103  arpslave2

创建binlog的目录

mkdir -p /data/mysqldata/binlog
chown -R mysql:mysql /data/mysqldata/binlog

101节点配置:

server-id = 200101

log-bin = /data/mysqldata/binlog/mysql-bin
binlog_format= row
max_binlog_size= 512m
relay-log = /data/mysqldata/binlog/relay-bin
expire-logs-days = 14

lower_case_table_names = 1
character-set-server = utf8
log_slave_updates = 1

102节点配置:

server-id = 200102

log-bin = /data/mysqldata/binlog/mysql-bin
binlog_format= row
max_binlog_size= 512m
relay-log = /data/mysqldata/binlog/relay-bin
expire-logs-days = 14

read_only = ON
relay_log_purge = 0

lower_case_table_names = 1
character-set-server = utf8
log_slave_updates = 1

103节点配置:

server-id = 200103

log-bin = /data/mysqldata/binlog/mysql-bin
binlog_format= row
max_binlog_size= 512m
relay-log = /data/mysqldata/binlog/relay-bin
read_only = ON
relay_log_purge = 0
expire-logs-days = 14

lower_case_table_names = 1
character-set-server = utf8
log_slave_updates = 1

3.1.3 配置一主多从

master节点配置:

MariaDB [(none)]>grant replication slave,replication client on *.* to 'repl'@'172.30.200.%' identified by 'repl7101';
MariaDB [(none)]> show master status;
+------------------+----------+--------------+------------------+
| File             | Position | Binlog_Do_DB | Binlog_Ignore_DB |
+------------------+----------+--------------+------------------+
| mysql-bin.000001 |      548 |              |                  |
+------------------+----------+--------------+------------------+
1 row in set (0.000 sec)

slave节点配置:

grant replication slave,replication client on *.* to 'repl'@'172.30.200.%' identified by 'repl7101';

change master to master_host='172.30.200.101',
master_user='repl',
master_password='repl7101',
master_log_file='mysql-bin.000001',
master_log_pos=548;

start slave;
show slave status\G;

至此，一主多从配置完毕。

3.2 MHA配置

3.2.1 master权限授予

可以在所有节点上面配置，其拥有管理权限，目前只需在master结点上设置权限:

grant all on *.* to 'mhaadmin'@'172.30.%.%' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpmgr' identified by 'mha7101';

3.2.2 ssh互信

四个节点都执行如下语句:

ssh-keygen -t rsa
ssh-copy-id -i .ssh/id_rsa.pub root@arpmgr

然后在arpmgr结点上面，可以看到authorized_keys文件的信息内容如下:

[root@localhost .ssh]# cat authorized_keys
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDY3yFhR5uzcFEpd+q+1Uw/cRF9ZRraygms7OZoefLFzY/ydSi6yYuCighG8WquvRep7XDNjFI71HAUagSoXiyPoCe1lqEnzpxSc+fQpIeQqEhUmLJ2bk+R83EskzwRGh+S/D4yp/swWz1vRgUGoTWevLCs33q7ZrsM8i+jB0uwZmzOV+CyQAPW9vLkRjZa4y1sx65lbR0HbdTQWQYZ4IyZauoU8XQjAIOs/CdLw2nBt8dPO53jT7NS7Ywx6eu/Wj9k/sYVVZT3jTb+pBIVs+Du5+tdUDX5aLKzxINpLlqNhorNevoC9iE0Ame1qvYonQfyWQ52Ae0y+58vFfG6PyV3 root@localhost.localdomain
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQC/ZPihYSC6ArawKRU75aQRVSFsQ5S89SrYHGWdzyluB4spj+UDUmWH1kLGYr715/HD5hh22KdLmIs7R4jviOeao1HK52fpMvklYaNtYRHuV63Zkg5sOLvLfhrHdta9wuHlW1NyWx75+wIl2LvKBRtnSddwf5ZvitJ/kChf2gpNhHAWidyjGsPoJdr0OBCNHvz1y6oON6cnMb07ExaIjptRnkbCOU0QSVjFq4+Jmh8zTTbJC2up50s15gSfWXH0+WLXmJXJGkvgHdSYqw4vJt/l25f5qAKKZsfnyfC0iyct4GyHPF6trpvQ/c2lqr/Rg4xLWgdxlyt4aBJYl5adIRK/ root@localhost.localdomain
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQDba26wV0KwQNTb4pKuiFDCcVMNRLGMXSiJC8ucN4/KIqzoOYJ747QL8GL5F8ePnRaZ1rtOwdjnlTiC0a4Tcg4JLs+JSnJgzvepuixmGgSJfLbJ36iN1WFh6fP2GZEDdR7Qum4sBUpQyYJ20Kf9rKfQQv2wq6csK5IlFk/OoO+zTySauLnYvRxvKY2avVDXPPFJvpqimKXn59MIAoJr6YEKvncbYyqvrSgUy7klZDys9IIjYcWfO7VKjQ5bwbHrrKtNbedME+KPQld7e8ZVL66Omik4Z6ip7DQEHRKWMmuBIpL99AgOOjPLbzJFWLUPOwvy3DtmEBnZ+0NVf/1obC11 root@localhost.localdomain
ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQCwrgZtGC31EgixeY4SVl4h64m1r8LdL3hM4Be2/I+6Xw7hCzZyKKTAFgz9W/ukfx6WmZwoqp1VO/7Jp6KO1FhYOi5u0q6J1KIObFNp+3E6cB2P0q39WqmZpQ9cNPYrbs9U2Ej0L0JwUtf/xLh334PaSlv/LcNy+p1dWya2OqsBeraiXZ4MgEBzcb+0twkpfpD327VgT/mRHPmA6fPRJOOJti1u4isHeotE4i13YIqQYfBfmbfiLdXKAvgI8FuTf0i91Re/FUBOgBfBcJbqIQNR0Nh5wZ/LvNxkstDQvypZIZwiK+wN+aZZOQ7jF/+997Z9QQleC9OOoHOJR7+fisLb root@localhost.localdomain

正好有四个形如ssh-rsa 相关 的密钥信息。

把如上的公钥信息，拷贝到其余四台服务器上面:

scp authorized_keys root@arpmaster:~/.ssh/
scp authorized_keys root@arpslave1:~/.ssh/
scp authorized_keys root@arpslave2:~/.ssh/

测试ssh是否可用

[root@localhost .ssh]# ssh arpmaster
[root@localhost ~]# ssh arpslave1
[root@localhost ~]# ssh arpslave2
[root@localhost ~]# ssh arpmgr

3.2.3 安装mha包

mha安装包分为两个，一个是node,另外一个是manager。

四个节点安装:mha4mysql-node-0.57-0.el7.centos.noarch.rpm

管理节点安装:mha4mysql-manager-0.57-0.el7.centos.noarch.rpm

在安装``mha4mysql-node-0.57-0.el7.centos.noarch.rpm过程中,有对perl-DBD-mysql,perl-DBI`前置依赖,安装步骤如下:

yum install perl-DBD-mysql perl-DBI

在安装``mha4mysql-manager-0.57-0.el7.centos.noarch.rpm过程中,有对perl`前置依赖,安装步骤如下:

安装yum 扩展包
yum install https://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpm

yum install perl-ExtUtils-CBuilder perl-ExtUtils-MakeMaker perl-Config-Tiny perl-Log-Dispatch-* perl-Parallel-ForkManager

然后安装信息，都成功，如下:

[root@localhost ~]# rpm -ivh mha4mysql-node-0.57-0.el7.centos.noarch.rpm
准备中...                          ################################# [100%]
正在升级/安装...
   1:mha4mysql-node-0.58-0.el7.centos ################################# [100%]
[root@localhost ~]# rpm -ivh mha4mysql-manager-0.57-0.el7.centos.noarch.rpm
准备中...                          ################################# [100%]
正在升级/安装...
   1:mha4mysql-manager-0.58-0.el7.cent################################# [100%]

0.58中有一个super_read_only不可用在mariadb，所以使用0.57版本。

3.2.4 MHA管理节点配置

[root@localhost ~]# cd /etc/mha_master/
[root@localhost mha_master]# vi /etc/mha_master/mha.cnf

配置文件内容如下:

[server default]
user=mhaadmin
password=mha7101
manager_workdir=/etc/mha_master/app1
manager_log=/etc/mha_master/manager.log
remote_workdir=/data/mha_master/app1
repl_user=repl
repl_password=repl7101
ping_interval=1
[server1]
hostname=172.30.200.101
ssh_port=22
[server2]
hostname=172.30.200.102
ssh_port=22
candidate_master=1
[server3]
hostname=172.30.200.103
ssh_port=22
no_master=1

3.2.5 MHA节点检测

1. 在管理节点检测ssh连通性如下:

   [root@localhost ~]# masterha_check_ssh -conf=/etc/mha_master/mha.cnf
   
   

   有如下日志,代表正常:

   Thu Jan  9 14:43:09 2020 - [warning] Global configuration file /etc/masterha_default.cnf not found. Skipping.
   Thu Jan  9 14:43:09 2020 - [info] Reading application default configuration from /etc/mha_master/mha.cnf..
   root@172.30.200.103(172.30.200.103:22) to root@172.30.200.101(172.30.200.101:22)..
   Thu Jan  9 14:43:11 2020 - [debug]   ok.
   Thu Jan  9 14:43:11 2020 - [debug]  Connecting via SSH from root@172.30.200.103(172.30.200.103:22) to root@172.30.200.102(172.30.200.102:22)..
   Thu Jan  9 14:43:11 2020 - [debug]   ok.
   Thu Jan  9 14:43:12 2020 - [info] All SSH connection tests passed successfully.
   

2. 检测MySQL replication是否正常

   masterha_check_repl --conf=/etc/mha_master/mha.cnf
   

   有如下日志，说明正常:

   Thu Jan  9 14:44:54 2020 - [info] Slaves settings check done.
   Thu Jan  9 14:44:54 2020 - [info]
   172.30.200.101(172.30.200.101:3306) (current master)
    +--172.30.200.102(172.30.200.102:3306)
    +--172.30.200.103(172.30.200.103:3306)
   
   Thu Jan  9 14:44:54 2020 - [info] Checking replication health on 172.30.200.102..
   Thu Jan  9 14:44:54 2020 - [info]  ok.
   Thu Jan  9 14:44:54 2020 - [info] Checking replication health on 172.30.200.103..
   Thu Jan  9 14:44:54 2020 - [info]  ok.
   Thu Jan  9 14:44:54 2020 - [warning] master_ip_failover_script is not defined.
   Thu Jan  9 14:44:54 2020 - [warning] shutdown_script is not defined.
   Thu Jan  9 14:44:54 2020 - [info] Got exit code 0 (Not master dead).
   
   MySQL Replication Health is OK.
   

3.2.6 MHA启动

1. 启动mha manager:

   nohup masterha_manager --conf=/etc/mha_master/mha.cnf &> /etc/mha_master/manager.log &
   

2. 检测master节点状态:

   [root@localhost ~]# masterha_check_status --conf=/etc/mha_master/mha.cnf
   mha (pid:31709) is running(0:PING_OK), master:172.30.200.101
   

   说明主数据库172.30.200.101启动正常。

3. 关闭mha manager:

   masterha_stop -conf=/etc/mha_master/mha.cnf
   

3.2.7 MHA模拟故障

1. master直接kill mysql节点

   [root@localhost ~]# ps -ef |grep mysql
   root     19864     1  0 08:51 ?        00:00:00 /bin/sh /usr/local/mysql/bin/mysqld_safe --datadir=/data/mysqldata --pid-file=/data/mysqldata/localhost.localdomain.pid
   mysql    19976 19864  0 08:51 ?        00:00:13 /usr/local/mysql/bin/mysqld --basedir=/usr/local/mysql --datadir=/data/mysqldata --plugin-dir=/usr/local/mysql/lib/plugin --user=mysql --log-error=/data/mysqldata/mysqld.log --pid-file=/data/mysqldata/localhost.localdomain.pid --socket=/tmp/mysql.sock
   root     22166 21525  0 14:55 pts/0    00:00:00 grep --color=auto mysql
   [root@localhost ~]# kill -9 19864 19976
   

2. MHA转移日志。

   [root@localhost ~]# tail -f /etc/mha_master/manager.log
   
   From:
   172.30.200.101(172.30.200.101:3306) (current master)
    +--172.30.200.102(172.30.200.102:3306)
    +--172.30.200.103(172.30.200.103:3306)
   
   To:
   172.30.200.102(172.30.200.102:3306) (new master)
    +--172.30.200.103(172.30.200.103:3306)
   
   Master 172.30.200.101(172.30.200.101:3306) is down!
   
   Check MHA Manager logs at localhost.localdomain:/etc/mha_master/manager.log for details.
   
   Started automated(non-interactive) failover.
   The latest slave 172.30.200.102(172.30.200.102:3306) has all relay logs for recovery.
   Selected 172.30.200.102(172.30.200.102:3306) as a new master.
   172.30.200.102(172.30.200.102:3306): OK: Applying all logs succeeded.
   172.30.200.103(172.30.200.103:3306): This host has the latest relay log events.
   Generating relay diff files from the latest slave succeeded.
   172.30.200.103(172.30.200.103:3306): OK: Applying all logs succeeded. Slave started, replicating from 172.30.200.102(172.30.200.102:3306)
   172.30.200.102(172.30.200.102:3306): Resetting slave info succeeded.
   Master failover to 172.30.200.102(172.30.200.102:3306) completed successfully.
   

   从上述日志来看，172.30.200.102已经成为了新的master，而172.30.200.103还是slave数据库。

3.2.8 修复已损坏的节点加入MHA中

由于这里是实验环境，可以不到处mysqldump的备份。如果是生产环境恢复，可以停掉slave的SQL thread，记住对应的pos的位置，然后备份出数据，保证数据一致性之后，同步数据，恢复损坏的结点。

change master to master_host='172.30.200.102',
master_user='repl',
master_password='repl7101',
master_log_file='mysql-bin.000003',
master_log_pos=401;

查看slave状态:

MariaDB [(none)]> start slave;

MariaDB [(none)]> show slave status\G;
*************************** 1. row ***************************
                Slave_IO_State: Waiting for master to send event
                   Master_Host: 172.30.200.102
                   Master_User: repl
                   Master_Port: 3306
                 Connect_Retry: 60
               Master_Log_File: mysql-bin.000003
           Read_Master_Log_Pos: 401
                Relay_Log_File: relay-bin.000002
                 Relay_Log_Pos: 555
         Relay_Master_Log_File: mysql-bin.000003
              Slave_IO_Running: Yes
             Slave_SQL_Running: Yes

再次启动，如下:

[root@localhost ~]# nohup masterha_manager --conf=/etc/mha_master/mha.cnf &> /etc/mha_master/manager.log &
[root@localhost ~]# masterha_check_status --conf=/etc/mha_master/mha.cnf
mha (pid:31905) is running(0:PING_OK), master:172.30.200.101

至此，MHA实验完毕。由于生产环境会用到VIP，后续会继续编写。

MHA 问题集锦

错误一

日志错误:

Thu Jan  9 11:31:36 2020 - [info]   Connecting to root@172.30.200.102(172.30.200.102:22)..
Can't exec "mysqlbinlog": 没有那个文件或目录 at /usr/share/perl5/vendor_perl/MHA/BinlogManager.pm line 106.
mysqlbinlog version command failed with rc 1:0, please verify PATH, LD_LIBRARY_PATH, and client options

解决方法:

ln -s /usr/local/mysql/bin/mysqlbinlog /usr/bin/mysqlbinlog

错误二

日志错误:

Checking if super_read_only is defined and turned on..DBI connect(';host=172.30.200.102;port=3306','mhaadmin',...) failed: Access denied for user 'mhaadmin'@'arpslave1' (using password: YES) at /usr/share/perl5/vendor_perl/MHA/SlaveUtil.pm line 239

解决方法:

manager节点，执行:

grant all on *.* to 'mhaadmin'@'arpmgr' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpmaster' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpslave1' identified by 'mha7101';
grant all on *.* to 'mhaadmin'@'arpslave2' identified by 'mha7101';

错误三

日志如下:

    Testing mysql connection and privileges..sh: mysql: 未找到命令
mysql command failed with rc 127:0!

解决方法:

ln -s /usr/local/mysql/bin/mysql /usr/local/bin/mysql