[MGR-Mysql的组复制集群]详细搭建部署过程

[MGR-Mysql]的组复制集群]详细搭建部署过程

 author:石鲁坤

1,关于MySQL Group Replication
 
基于组的复制(Group-basedReplication)是一种被使用在容错系统中的技术。Replication-group(复制组)是由能够相互通信的多个服务器(节点)组成的。
 
在通信层,Groupreplication实现了一系列的机制:比如原子消息(atomicmessage delivery)和全序化消息(totalorderingof messages)。
 
这些原子化,抽象化的机制,为实现更先进的数据库复制方案提供了强有力的支持。
  
MySQL Group Replication正是基于这些技术和概念,实现了一种多主全更新的复制协议。

简而言之,一个Replication-group就是一组节点,每个节点都可以独立执行事务,而读写事务则会在于group内的其他节点进行协调之后再commit。
 
因此,当一个事务准备提交时,会自动在group内进行原子性的广播,告知其他节点变更了什么内容/执行了什么事务。
 
这种原子广播的方式,使得这个事务在每一个节点上都保持着同样顺序。
 
这意味着每一个节点都以同样的顺序,接收到了同样的事务日志,所以每一个节点以同样的顺序重演了这些事务日志,最终整个group保持了完全一致的状态。
 
然而,不同的节点上执行的事务之间有可能存在资源争用。这种现象容易出现在两个不同的并发事务上。
 
假设在不同的节点上有两个并发事务,更新了同一行数据,那么就会发生资源争用。
 
面对这种情况,GroupReplication判定先提交的事务为有效事务,会在整个group里面重演,后提交的事务会直接中断,或者回滚,最后丢弃掉。
 
因此,这也是一个无共享的复制方案,每一个节点都保存了完整的数据副本。这个复制方案,在某种程度上,和数据库状态机(DBSM)的Replication方法比较类似。
 
 MGR的限制
仅支持InnoDB表,并且每张表一定要有一个主键,用于做write set的冲突检测;
必须打开GTID特性,二进制日志格式必须设置为ROW,用于选主与write set
COMMIT可能会导致失败,类似于快照事务隔离级别的失败场景
目前一个MGR集群最多支持9个节点
不支持外键于save point特性,无法做全局间的约束检测与部分部分回滚
二进制日志不支持binlog event checksum
 

2,安装mysql 5.7.18
 
在三台db服务器上面设置/etc/hosts映射,如下:
 
192.168.117.129  mgr1
192.168.117.130  mgr2
192.168.117.131  mgr3
 
安装的数据库服务器:

在每个节点上执行安装:

1、下载二进制软件包:
  mysql-5.7.18-linux-glibc2.5-x86_64.tar.gz

2、添加MySQL用户和组:
  [root@localhost ~]# userdel mysql  (因为是新系统,所以就把现有的MySQL用户给删除了)
  [root@localhost ~]# groupadd mysql
  [root@localhost ~]# useradd -g mysql mysql
  [root@localhost ~]# passwd mysql     (MySQL密码要设置的复杂一点,或者在创建用户的时候设置不让登陆系统)

3、创建安装MySQL目标,并解压软件到指定目录
  [root@localhost ~]# mkdir -p /data/mysql/db_file  在mysql 下创建bin_log  relay_log(从节点) log 目录
  [root@localhost ~]# chown -R mysql:mysql /data
  [root@localhost ~]# tar -xvf  mysql-5.7.18-linux-glibc2.5-x86_64 
  [root@localhost ~]# mv mysql-5.7.18-linux-glibc2.5-x86_64/ /usr/local/mysql
  [root@localhost ~]# chown -R mysql:mysql /usr/local/mysql

4、切换MySQL用户,并初始化数据库
  [root@localhost ~]# su - mysql
  [mysql@localhost ~]$/usr/local/mysql/bin/mysqld --initialize --user=mysql --basedir=/usr/local/mysql/ --datadir=/data/mysql/db_file/

my.cnf的详细配置:

[root@localhost ~]# cat /etc/my.cnf 
[client]
port = 3306
socket = /tmp/mysql.sock
 
[mysqld]
port = 3306
socket = /tmp/mysql.sock
back_log = 80
basedir = /usr/local/mysql
datadir = /data/mysql/db_file
#skip-grant-tables
#-------------------gobal variables------------#
gtid_mode = ON
enforce_gtid_consistency = ON
master_info_repository = TABLE
relay_log_info_repository = TABLE
binlog_checksum = NONE
log_slave_updates = ON
log-bin = /usr/local/mysql/log/mysql-bin
transaction_write_set_extraction = XXHASH64
loose-group_replication_group_name = 'ce9be252-2b71-11e6-b8f4-00212844f856'
loose-group_replication_start_on_boot = off
loose-group_replication_local_address = '192.168.117.129:33061'
loose-group_replication_group_seeds ='192.168.117.129:33061,192.168.117.130:33061,192.168.117.131:33061'
loose-group_replication_bootstrap_group = off
loose-group_replication_single_primary_mode = true (单主:true 多主:false)
loose-group_replication_enforce_update_everywhere_checks = false
max_connect_errors = 20000
max_connections = 2000
wait_timeout = 3600
interactive_timeout = 3600
net_read_timeout = 3600
net_write_timeout = 3600
table_open_cache = 1024
table_definition_cache = 1024
thread_cache_size = 512
open_files_limit = 10000
character-set-server = utf8
collation-server = utf8_bin
skip_external_locking
performance_schema = 1
user = mysql
myisam_recover_options = DEFAULT
skip-name-resolve
local_infile = 0
lower_case_table_names = 0
 
#--------------------innoDB------------#
innodb_buffer_pool_size = 512M
innodb_data_file_path = ibdata1:200M:autoextend
innodb_flush_log_at_trx_commit = 1
innodb_io_capacity = 600
innodb_lock_wait_timeout = 120
innodb_log_buffer_size = 8M
innodb_log_file_size = 128M
innodb_log_files_in_group = 3
innodb_max_dirty_pages_pct = 85
innodb_read_io_threads = 8
innodb_write_io_threads = 8
innodb_support_xa = 1
innodb_thread_concurrency = 32
innodb_file_per_table
innodb_rollback_on_timeout
 
#------------session variables-------#
join_buffer_size = 8M
key_buffer_size = 256M
bulk_insert_buffer_size = 8M
max_heap_table_size = 96M
tmp_table_size = 96M
read_buffer_size = 8M
sort_buffer_size = 2M
max_allowed_packet = 64M
read_rnd_buffer_size = 32M
 
#------------MySQL Log----------------#
log_bin = /data/mysql/bin_log/mysql-bin3306
binlog_format = row
sync_binlog = 1
expire_logs_days = 15
max_binlog_cache_size = 128M
max_binlog_size = 256M
binlog_cache_size = 64k
slow_query_log
log-slow-admin-statements
log_warnings = 1
long_query_time = 2
 
#---------------replicate--------------#
relay_log = /data/mysql/bin_log/relay_bin
server-id = 1
init_slave = 'set sql_mode=STRICT_ALL_TABLES'
log-slave-updates
[myisamchk]
key_buffer = 512M
sort_buffer_size = 512M
read_buffer = 8M
write_buffer = 8M
[mysqlhotcopy]
interactive-timeout
[mysqld_safe]
open-files-limit = 8192
log-error = /data/mysql/log/mysqld_error.log


复制代码
注意每个节点的server_id、loose-group_replication_local_address等都配置成自己的相应参数
 


[mysql@localhost ~]$ /usr/local/mysql/support-files/mysql.server start 
[mysql@localhost ~]$ /usr/local/mysql/bin/mysqld_safe --defaults-file=/ect/my.cnf  --user=mysql & (效果一样)

6、进入数据库检查:
[mysql@localhost ~]$  /usr/local/mysql/bin/mysql -u root -p 
Enter password: 
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 7
Server version: 5.7.18-log MySQL Community Server (GPL)

Copyright (c) 2000, 2017, Oracle and/or its affiliates. All rights reserved.

Oracle is a registered trademark of Oracle Corporation and/or its
affiliates. Other names may be trademarks of their respective
owners.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql> set password='root';  (因为刚才使用的密码是个默认的登录密码,所以我们要修改)
mysql> flush privileges;

然后再重新登录MySQL即可,进行检查3,创建复制环境
 
在mgr1/mgr2/mgr3上建立复制账号:
 
mysql>  GRANT REPLICATION SLAVE ON *.* TO 'repl'@'192.168.%' IDENTIFIED BY 'repl';   
 
 
4,安装group replication插件
 
在mgr1、mgr2、mgr3上依次安装group replication插件
mysql> INSTALL PLUGIN group_replication SONAME 'group_replication.so';  
 
 
5,配置group replication参数
 
确保binlog_format是row格式。
mysql> show variables like 'binlog_format';                                              
+---------------+-------+  
| Variable_name | Value |  
+---------------+-------+  
| binlog_format | ROW   |  
+---------------+-------+  
1 row in set (0.00 sec)  
   

配置文件配置

(1)  mgr1上的my.cnf配置:

server-id=1

transaction_write_set_extraction = XXHASH64
loose-group_replication_group_name = 'ce9be252-2b71-11e6-b8f4-00212844f856'
loose-group_replication_start_on_boot = off (单主时作为主节点,建议修改为on,以后重启不用在set)
loose-group_replication_local_address = '192.168.117.129:33061'
loose-group_replication_group_seeds ='192.168.117.129:33061,192.168.117.130:33061,192.168.117.131:33061'
loose-group_replication_bootstrap_group = off
loose-group_replication_single_primary_mode = true (单主:true 多主:false)
loose-group_replication_enforce_update_everywhere_checks = false

 
(2)mgr2上的my.cnf配置:

server-id=2
transaction_write_set_extraction = XXHASH64
loose-group_replication_group_name = "ce9be252-2b71-11e6-b8f4-00212844f856"
loose-group_replication_start_on_boot = off

loose-group_replication_local_address = '192.168.117.130:33061'
loose-group_replication_group_seeds ='192.168.117.129:33061,192.168.117.130:33061,192.168.117.131:33061'
loose-group_replication_bootstrap_group = off
loose-group_replication_single_primary_mode = true (单主:true 多主:false)
loose-group_replication_enforce_update_everywhere_checks = false


(3)mgr3上的my.cnf配置:

server-id=3
transaction_write_set_extraction = XXHASH64
loose-group_replication_group_name = "ce9be252-2b71-11e6-b8f4-00212844f856"
loose-group_replication_start_on_boot = off

loose-group_replication_local_address = '192.168.117.131:33061'
loose-group_replication_group_seeds ='192.168.117.129:33061,192.168.117.130:33061,192.168.117.131:33061'
loose-group_replication_bootstrap_group = off
loose-group_replication_single_primary_mode = true (单主:true 多主:false)
loose-group_replication_enforce_update_everywhere_checks = false
确认配置后重启集群

6,启动mgr集群
 
开始构建group replication集群,通常操作命令
在mgr1、mgr2、mgr3上依次执行
mysql>  CHANGE MASTER TO MASTER_USER='repl', MASTER_PASSWORD='repl' FOR CHANNEL 'group_replication_recovery';  
Query OK, 0 rows affected, 2 warnings (0.02 sec)  
 
mgr1上建立基本主库master库:
# 设置group_replication_bootstrap_group为ON是为了标示以后加入集群的服务器以这台服务器为基准,以后加入的就不需要设置。  
mysql> SET GLOBAL group_replication_bootstrap_group = ON;  
Query OK, 0 rows affected (0.00 sec)  
   
mysql>  START GROUP_REPLICATION;  
Query OK, 0 rows affected (1.03 sec)  
   
mysql> select * from performance_schema.replication_group_members; 

 
 
mgr2上启动group_replication: 
mgr2上mysql命令行上执行启动:
mysql> set global group_replication_allow_local_disjoint_gtids_join=ON; (该参数也可以直接写入my.cnf)
 
mysql> start group_replication;
 
mysql> select * from performance_schema.replication_group_members;

 
mgr3上启动group_replication:
-- mgr3命令行上执行:  
mysql> set global group_replication_allow_local_disjoint_gtids_join=ON;  (该参数也可以直接写入my.cnf)
   
mysql> start group_replication;  
   
-- 再去master库mgr1上,查看group_replication成员,会有mgr3的显示,而且已经是ONLINE了  
mysql> select * from performance_schema.replication_group_members;

   
 
最后查看集群状态,都为ONLINE就表示OK:
mysql> select * from performance_schema.replication_group_members;  

    
 
7,验证集群复制功能
 
查看那个是主节点
mysql> select variable_value from performance_schema.global_status where variable_name="group_replication_primary_member";

 
在master库mgr1上建立测试库mgr1,测试表t1,录入一条数据
mysql> create database mgr1;  
Query OK, 1 row affected (0.00 sec)  
   
mysql> create table mgr1.t1(id int,cn varchar(32));  
Query OK, 0 rows affected (0.02 sec)  
   
mysql> insert into t1(id,cn)values(1,'a');  
ERROR 3098 (HY000): The table does notcomply with the requirements by an external plugin.   
-- # 这里原因是group_replaction环境下面,表必须有主键不然不允许往里insert值。所以修改表t1,将id字段设置程主键即可。  
mysql> alter table t1 add primary key(id); 
   
mysql> insert into t1(id,cn)values(1,'a');    
 
 
去mgr2/mgr3上可以看到数据已经同步过去
mysql> select * from mgr1.t1;  
+----+------+  
| id | cn  |  
+----+------+  
|  1| a    |  
+----+------+  
1 row in set (0.00 sec)  
   
 
然后在mgr2/mgr3上执行inert操作,则拒绝,因为mgr2、mgr3为readonly
mysql> insert into t1 select 2,'b';  
ERROR 1290 (HY000): The MySQL server isrunning with the --super-read-only option so it cannot execute this statement  
 
 
8,问题记录
 
8.1问题记录一
 
MySQL窗口报错:
 
ERROR 3092 (HY000): The server is notconfigured properly to be an active member of the group. Please see moredetails on error log.
 
后台ERROR LOG报错:
 
[ERROR] Plugin group_replication reported:'This member has more executed transactions than those present in the group.Local transactions: f16f7f74-c283-11e6-ae37-fa163ee40410:1 > Grouptransactions: 3c992270-c282-11e6-93bf-fa163ee40410:1,
 
 aaaaaa:1-5'
 
 [ERROR]Plugin group_replication reported: 'The member contains transactions notpresent in the group. The member will now exit the group.'
 
 [Note] Plugin group_replication reported: 'Toforce this member into the group you can use the group_replication_allow_local_disjoint_gtids_joinoption'
 
 
 
【解决办法】:
 
根据提示打开group_replication_allow_local_disjoint_gtids_join选项,mysql命令行执行:
 
mysql> set globalgroup_replication_allow_local_disjoint_gtids_join=ON;
 
 
 
再执行开启组复制:
 
mysql> start group_replication;
 
Query OK, 0 rows affected (7.89 sec)
 
 
8.2 问题记录二RECOVERING  
 
在mgr1上查询集群组成员
mysql> SELECT * FROM performance_schema.replication_group_members;  
+---------------------------+--------------------------------------+----------------------+-------------+--------------+  
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST          | MEMBER_PORT | MEMBER_STATE |  
+---------------------------+--------------------------------------+----------------------+-------------+--------------+  
| group_replication_applier | 3d872c2e-d670-11e6-ac1f-b8ca3af6e36c  |  |         3306 | ONLINE        
| group_replication_applier | ef8ac2de-d671-11e6-9ba4-18a99b763071 |  |        3306 | RECOVERING   |  
| group_replication_applier | fdf2b02e-d66f-11e6-98a8-18a99b76310d  |  |        3306 | RECOVERING   |  
+---------------------------+--------------------------------------+----------------------+-------------+--------------+  
3 rows in set (0.00 sec) 
 
 
【解决办法】:
 
看报错[ERROR] Slave I/O for channel'group_replication_recovery': error connecting to master 'repl@:3306'- retry-time: 60  retries: 1, Error_code:2005,连接master库不上,所以问题在这里,我们赋予的复制账号是ip的repl@'192.168.%',所以还需要做一个hostname()和mgr1的ip地址192.168.1.20的映射关系。
 
 
 
建立hostname和ip映射
vim /etc/hosts  
192.168.1.129 mgr1                                             
192.168.1.130 mgr2   
192.168.1.131 mgr3   
 
然后在mgr2上执行如下命令后重新开启group_replication即可。
mysql> stop group_replication;  
Query OK, 0 rows affected (0.02 sec)                                                   
   
mysql> start group_replication;  
Query OK, 0 rows affected (5.68 sec)  
   
 
再去master库mgr1上,查看group_replication成员,会有mgr2的显示
mysql>  SELECT * FROM performance_schema.replication_group_members;  
+---------------------------+--------------------------------------+----------------------+-------------+--------------+  
| CHANNEL_NAME              | MEMBER_ID                            | MEMBER_HOST          | MEMBER_PORT | MEMBER_STATE 
+---------------------------+--------------------------------------+----------------------+-------------+--------------+  
| group_replication_applier | 3d872c2e-d670-11e6-ac1f-b8ca3af6e36c |  |        3306 | ONLINE        
| group_replication_applier | fdf2b02e-d66f-11e6-98a8-18a99b76310d |           |        3306 | ONLINE       |  
+---------------------------+--------------------------------------+----------------------+-------------+--------------+  
2 rows in set (0.00 sec)  
   
 
 
8.3问题记录三
 
操作问题
 
 
mysql> START GROUP_REPLICATION;
 
ERROR 3092 (HY000): The server is notconfigured properly to be an active member of the group. Please see moredetails on error log.
 
 
【解决办法】:
 
mysql> SET GLOBALgroup_replication_bootstrap_group = ON;
 
mysql> START GROUP_REPLICATION;
 
9,日常维护步骤:
 
1、如果从库某一节点关闭
 
    stop group_replication;
 
2、如果所有的库都关闭后,第一个库作为主库首先执行
 
    set global group_replication_bootstrap_group=ON;
 
    start group_replication;
 
剩下的库直接执行即可!
 
    start group_replication;
 
3、如果主库故障,会自动从两个从库选出一个主库,主库启动后再次执行如下命令后会变成从库
 
    start group_replication;


重点参数详解

 

MGR一定要开启GTID功能gtid_mode = ON,enforce_gtid_consistency = ON

MGR在当前MySQL版本中,必须要设置binlog_checksum=none,并且binlog_format=row

MGR需要多源复制功能,所以需要将主从库之间的信息记录到表中。设置master_info_repository = TABLE和relay_log_info_repository = TABLE

MGR中开启主键信息采集功能:

transaction_write_set_extraction= XXHASH64

MGR中需要设置组的名字,该group_replication_group_name是通过select uuid()来生成的。

MGR中本地成员的地址由该参数group_replication_local_address决定,本例中当前节点是node2。后面的端口是每个节点都需要一个独立的TCP端口号,节点之间通过这个端口号进行通信。

MGR中其他成员的地址由这个参数group_replication_group_seeds决定

MGR采取多主模式时,需要让该参数:

group_replication_single_primary_mode=off(默认是单主模式)。

于此同时,还需要将该参数:group_replication_enforce_update_everywhere_checks=true,目的则是为了做多主模式限制检测。

 

注:加载到参数文件里面,需要在每个参数的前面加loose。

 

这里再强调一个参数:group_replication_auto_increment_increment

该参数代表自增属性,默认值为7。我们要保证每个成员的该值相同,并且建议该参数的设置尽量比组内成员的个数大一些,方便后期集群的扩展。

该文章是我在参考网上资料后,自己部署时实际情况,优化了参数

如有问题探讨,请加QQ群:695182221

你可能感兴趣的:(数据库)