group replication是一种全新的高可用,高扩张的MySQL集群服务。
高一致性,基于原生复制及paxos协议的组复制技术,以插件方式提供一致数据安全保证;
高容错性,大多数服务正常就可继续工作,自动不同节点检测资源征用冲突,按顺序优先处理,内置自动防脑裂机制;
高扩展性,自动添加移除节点,并更新组信息;
高灵活性,单主模式和多主模式。单主模式自动选主,所有更新操作在主进行;多主模式,所有server同时更新。

0,修改hosts
vim /etc/hosts
localhost 127.0.0.1

1,实验环境
M1:127.0.0.1 3307
M2:127.0.0.1 3308

2,安装mysql
/usr/local/mysql57/bin/mysqld --initialize-insecure --user=dba --basedir=/usr/local/mysql57 --datadir=/data1/mysql3307
/usr/local/mysql57/bin/mysqld --initialize-insecure --user=dba --basedir=/usr/local/mysql57 --datadir=/data1/mysql3308

3,配置
M1:
[mysqld]

# GENERAL con#
user = dba
port = 3307
default_storage_engine = InnoDB
socket = /tmp/mysql3307.sock
pid_file = /data1/mysql3307/mysql.pid

# SAFETY #
max_allowed_packet = 64M
max_connect_errors = 1000000

# DATA STORAGE #binlog-format
datadir = /data1/mysql3307/

# BINARY LOGGING #
log_bin = /data1/mysql3307/3307-binlog
expire_logs_days = 10
#sync_binlog = 1
relay-log= /data1/mysql3307/3307-relaylog
#replicate-wild-do-table=hostility_url.%
#replicate-wild-do-table=guards.%

# CACHES AND LIMITS #
tmp_table_size = 32M
max_heap_table_size = 32M
query_cache_type = 1
query_cache_size = 0
max_connections = 5000
#max_user_connections = 200
thread_cache_size = 512
open_files_limit = 65535
table_definition_cache = 4096
table_open_cache = 4096
wait_timeout=7500
interactive_timeout=7500
binlog-format=row
character-set-server=utf8
skip-name-resolve
skip-character-set-client-handshake
back_log=1024

# INNODB #
#innodb_flush_method = O_DIRECT
innodb_data_home_dir = /data1/mysql3307/
#innodb_data_file_path = ibdata1:1G:autoextend
innodb_log_group_home_dir=/data1/mysql3307/
innodb_log_files_in_group = 3
innodb_log_file_size = 1G
innodb_flush_log_at_trx_commit = 2
innodb_file_per_table = 1
innodb_file_format=Barracuda
innodb_support_xa=0

innodb_io_capacity=500
innodb_max_dirty_pages_pct=90
innodb_read_io_threads=16
innodb_write_io_threads=8
innodb_buffer_pool_instances=4
innodb_thread_concurrency=0

#GTID

#gtid_mode = on
#enforce_gtid_consistency = on

# LOGGING #
log_error = /data1/mysql3307/error.log
#log_queries_not_using_indexes = 1
slow_query_log = 1
slow_query_log_file = /data1/mysql3307/mysql-slow.log
long_query_time=0.05
server_id=3307
innodb_buffer_pool_size = 1G
report-host=localhost #此处的localhost一定要存在与hosts文件里,即:localhost 127.0.0.1
report-port=3307
log_slave_updates = on

#group replication
gtid_mode = on
enforce_gtid_consistency = on
master_info_repository = table
relay_log_info_repository = table
binlog_checksum=NONE
log_slave_updates=ON

transaction_write_set_extraction=XXHASH64
loose-group_replication_group_name="aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa"
loose-group_replication_start_on_boot=off
loose-group_replication_local_address= "localhost:24901"

#该参数可动态修改: set global group_replication_group_seeds="localhost:24901,localhost:24902,localhost:24093",后续写入配置文件
loose-group_replication_group_seeds= "localhost:24901,localhost:24902" 
loose-group_replication_single_primary_mode = true
loose-group_replication_bootstrap_group= off
loose-group_replication_enforce_update_everywhere_checks = false

### mysql_version=mysql57
[mysql]
prompt = \u@\h:\p [\d]>

4,启动mysql
nohup /usr/local/mysql57/bin/mysqld_safe --defaults-file=/data1/mysql3307/my3307.cnf 2>/dev/null &
mysql -uroot -S /tmp/mysql3307.sock -p123456

set sql_log_bin=0;
create user rpl_user@'%';
grant replication slave on *.* to rpl_user@'%' identified by 'rpl_pass';
flush privileges;
set sql_log_bin=1;
change master to master_user='rpl_user',master_password='rpl_pass' for channel 'group_replication_recovery';

5,启动group replication
install plugin group_replication soname 'group_replication.so';
show plugins;
set global group_replication_bootstrap_group=on; //只需在一个实例上执行
start group_replication;
set global group_replication_bootstrap_group=off;
select * from performance_schema.replication_group_members;

6,测试(创建一个数据库,新建一个实例加入到group replication中后会自动同步到新实例)
create database test;
use test;
create table t(id int primary key auto_increment,name text);
insert into t values(1,'Lucas');
show binlog evnets;

7,添加一个新实例
M2:

[mysqld]

# GENERAL con#
user = dba
port = 3308
default_storage_engine = InnoDB
socket = /tmp/mysql3308.sock
pid_file = /data1/mysql3308/mysql.pid

# SAFETY #
max_allowed_packet = 64M
max_connect_errors = 1000000

# DATA STORAGE #binlog-format
datadir = /data1/mysql3308/

# BINARY LOGGING #
log_bin = /data1/mysql3308/3308-binlog
expire_logs_days = 10
#sync_binlog = 1
relay-log= /data1/mysql3308/3308relaylog
#replicate-wild-do-table=hostility_url.%
#replicate-wild-do-table=guards.%

# CACHES AND LIMITS #
tmp_table_size = 32M
max_heap_table_size = 32M
query_cache_type = 1
query_cache_size = 0
max_connections = 5000
#max_user_connections = 200
thread_cache_size = 512
open_files_limit = 65535
table_definition_cache = 4096
table_open_cache = 4096
wait_timeout=7500
interactive_timeout=7500
binlog-format=row
character-set-server=utf8
skip-name-resolve
skip-character-set-client-handshake
back_log=1024

# INNODB #
#innodb_flush_method = O_DIRECT
innodb_data_home_dir = /data1/mysql3308/
#innodb_data_file_path = ibdata1:1G:autoextend
innodb_log_group_home_dir=/data1/mysql3308/
innodb_log_files_in_group = 3
innodb_log_file_size = 1G
innodb_flush_log_at_trx_commit = 2
innodb_file_per_table = 1
innodb_file_format=Barracuda
innodb_support_xa=0

innodb_io_capacity=500
innodb_max_dirty_pages_pct=90
innodb_read_io_threads=16
innodb_write_io_threads=8
innodb_buffer_pool_instances=4
innodb_thread_concurrency=0

#GTID

#gtid_mode = on
#enforce_gtid_consistency = on

# LOGGING #
log_error = /data1/mysql3308/error.log
#log_queries_not_using_indexes = 1
slow_query_log = 1
slow_query_log_file = /data1/mysql3308/mysql-slow.log
long_query_time=0.05
server_id=3307
innodb_buffer_pool_size = 1G
report-host=localhost #此处的localhost一定要存在与hosts文件里,即:localhost 127.0.0.1
report-port=3307
log_slave_updates = on

#group replication
gtid_mode = on
enforce_gtid_consistency = on
master_info_repository = table
relay_log_info_repository = table
binlog_checksum=NONE
log_slave_updates=ON

transaction_write_set_extraction=XXHASH64
loose-group_replication_group_name="aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa"
loose-group_replication_start_on_boot=off
loose-group_replication_local_address= "localhost:24901"
loose-group_replication_group_seeds= "localhost:24901,localhost:24902"
loose-group_replication_single_primary_mode = true
loose-group_replication_bootstrap_group= off
loose-group_replication_enforce_update_everywhere_checks = false

### mysql_version=mysql57
[mysql]
prompt = \u@\h:\p [\d]>

8,启动实例 & 配置group replication
nohup /usr/local/mysql57/bin/mysqld_safe --defaults-file=/data1/mysql3308/my3308.cnf 2>/dev/null &
mysql -uroot -S /tmp/mysql3308.sock -p123456

set sql_log_bin=0;
create user rpl_user@'%';
grant replication slave on *.* to rpl_user@'%' identified by 'rpl_pass';
flush privileges;
set sql_log_bin=1;
change master to master_user='rpl_user',master_password='rpl_pass' for channel 'group_replication_recovery';

9,启动group replication
install plugin group_replication soname 'group_replication.so';
show plugins;

set global group_replication_allow_local_disjoint_gtids_join=ON;

start group_replication;

select * from performance_schema.replication_group_members;

10,测试同步是否正常
/usr/local/mysql57/bin/mysql -uroot -p123456 -S /tmp/mysql3308.sock

show databases;

注意:之前在新加实例的过程中新实例一直报recovering,主要是mysql的配置中report-host未配置,或者是report-host配置的值不在系统hosts里面

默认情况下:只有最开始配置的那台instance会设置super_read_only=off,即可以写入数据,其他后加入的都会默认为super_read_only=on,即只能读。

group replication分single-primary和multi-primary两种模式。

multi-primary模式不支持SERIALIZABLE隔离级别和外键约束。

这些限制可以通过 group_replication_enforce_update_everywhere_checks 参数来开启或关闭,single-primary模式下,该参数必须为OFF。该参数只能在group replication停止的情况下修改

下面介绍这两种模式

1,single primary

该模式下,只能有一个实例可接受写操作,组内其他成员会自动被设置为只读模式(super_read_only=on),若组内原主offline,会自动提升组内一个成员为primary,并设置super_read_only=off,接受读写操作,后续offline的主再加入进来后,会被设置成只读模式。整个切换过程全自动,无需人工干预。

选主规则:根据group中UUID字典排序;当主挂掉后,根据UUID 字典排序,选第排在最前面的UUID做新主。

2,multi primary

single primary 模式下,若原主挂掉,需要找出新主,如何确认primary?

SELECT VARIABLE_VALUE FROM performance_schema.global_status WHERE VARIABLE_NAME= 'group_replication_primary_member';

MySQL Group Replication的更多相关文章

  1. MySQL Group Replication 技术点

    mysql group replication,组复制,提供了多写(multi-master update)的特性,增强了原有的mysql的高可用架构.mysql group replication基 ...

  2. MySQL Group Replication 动态添加成员节点

    前提: MySQL GR 3节点(node1.node2.node3)部署成功,模式定为多主模式,单主模式也是一样的处理. 在线修改已有GR节点配置 分别登陆node1.node2.node3,执行以 ...

  3. Docker Images for MySQL Group Replication 5.7.14

    In this post, I will point you to Docker images for MySQL Group Replication testing. There is a new ...

  4. Percona XtraDB Cluster vs Galera Cluster vs MySQL Group Replication

    Percona XtraDB Cluster vs Galera Cluster vs MySQL Group Replication Overview Galera Cluster 由 Coders ...

  5. Mysql 5.7 基于组复制(MySQL Group Replication) - 运维小结

    之前介绍了Mysq主从同步的异步复制(默认模式).半同步复制.基于GTID复制.基于组提交和并行复制 (解决同步延迟),下面简单说下Mysql基于组复制(MySQL Group Replication ...

  6. MySQL group replication介绍

    “MySQL group replication” group replication是MySQL官方开发的一个开源插件,是实现MySQL高可用集群的一个工具.第一个GA版本正式发布于MySQL5.7 ...

  7. mysql group replication 主节点宕机恢复

    一.mysql group replication 生来就要面对两个问题: 一.主节点宕机如何恢复. 二.多数节点离线的情况下.余下节点如何继续承载业务. 在这里我们只讨论第一个问题.也就是说当主结点 ...

  8. mysql group replication观点及实践

    一:个人看法 Mysql  Group Replication  随着5.7发布3年了.作为技术爱好者.mgr 是继 oracle database rac 之后. 又一个“真正” 的群集,怎么做到“ ...

  9. MySQL Group Replication配置

    MySQL Group Replication简述 MySQL 组复制实现了基于复制协议的多主更新(单主模式). 复制组由多个 server成员构成,并且组中的每个 server 成员可以独立地执行事 ...

  10. Mysql Group Replication 简介及单主模式组复制配置【转】

    一 Mysql Group Replication简介    Mysql Group Replication(MGR)是一个全新的高可用和高扩张的MySQL集群服务.    高一致性,基于原生复制及p ...

随机推荐

  1. USACO 铂金 T1

    题意 给出一个数轴,每次可以选择停下并得到当前点的收益,或者继续随机向左右游走,走到边界游戏结束收益为0. 求从每个点出发的最大期望收益.(n<=1e5) 有一个显然的dp方程 这个方程是带环的 ...

  2. splay板子

    1, splay的一些基本操作. 使用前要插入$-INF,+INF$保证每个点的前驱后继存在. $get$函数在$x$存在时, 调用后, 根为$x$, 否则根为$x$的前驱或后继 const int ...

  3. python-day5笔记

    一.python基础--基本数据类型 (无论用户输入什么内容,input 都会存成字符串格式) 1.基本数据类型 1)数字类型 整型(整数)int:年级,年纪,等级,身份证号,QQ号,手机号,leve ...

  4. ehcache.xml详解

    <?xml version="1.0" encoding="UTF-8"?> <ehcache> <!-- Sets the pa ...

  5. 在linux中,如何增加、修改、删除、暂停和冻结用户名

    在Linux中,如何增加.修改.删除.暂停和冻结用户名 在操作增加.修改和删除用户名前,先认识linux中两个最重要的文件,它们就是账号管理最重要文件“/etc/passwd”与“etc/shadow ...

  6. spark shuffle内在原理说明

    在MapReduce框架中,shuffle是连接Map和Reduce之间的桥梁,Map的输出要用到Reduce中必须经过shuffle这个环节,shuffle的性能高低直接影响了整个程序的性能和吞吐量 ...

  7. 069——VUE中vuex之使用getters高效获取购物车商品总价

    <!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8&quo ...

  8. POJ 1008 简单模拟题

    e.... 虽然这是一道灰常简单的模拟题.但是米做的时候没有读懂第二个日历的计时方法.然后捏.敲完之后华丽的WA了进一个点.坑点就在一年的最后一天你是该输出本年的.e ...但是我好想并没有..看di ...

  9. zoj3732&& hdu4797 Graph Reconstruction

    Graph Reconstruction Time Limit: 2 Seconds      Memory Limit: 65536 KB      Special Judge Let there ...

  10. gitlba的搭建与使用

    实验环境继续使用git的实验环境,详情请点击连接https://www.cnblogs.com/cash-su/p/10131632.html 首先给服务器做一个本机的映射 [root@git1 ~] ...