Mysql 8.0 --MGR多主+Atlas代理（提一下Mysql-router代理）

本文亦有参考网上其他资料，篇数较多不一一列出，在此提前一并致谢。

壹、首先MGR多主搭建过程：

环境准备：

VMware

CentOS Linux release 7.3.1611 (Core)

一台虚拟机，IP分配：192.168.1.78；端口使用3306
按照Mysql8.0的单机安装方式进行安装；

然后将该机复制两台，
分别修改IP为192.168.1.78和192.168.1.78，以及两台新机器配置文件的serverid；
vim /etc/hosts
192.168.1.78 mgr-mul1
192.168.1.79 mgr-mul2
192.168.1.80 mgr-mul3

一、单机安装(192.168.1.78)
修改系统的打开文件限制（选做）：
vim /etc/security/limits.conf
* soft nofile 65535
* hard nofile 65535

配置文件：
vim /etc/my.cnf
-----------------------------------------------------------------------
[client]
port = 3306
socket = /tmp/mysql3306.sock
default_character_set = utf8mb4

[mysql]
default_character_set = utf8mb4

[mysqld]
lower_case_table_names     = 1
port = 3306
basedir = /usr/local/mysql8
datadir = /data/mysql3306/mysql3306
socket = /tmp/mysql3306.sock
pid_file = /data/mysql3306/mysql3306.pid
tmpdir = /tmp/
default_authentication_plugin = mysql_native_password
skip_name_resolve = 1
character_set_server = utf8mb4
collation_server = utf8mb4_unicode_ci
max_connections = 2000
max_connect_errors = 10000
interactive_timeout = 600
wait_timeout = 600
table_open_cache = 2048
thread_cache_size = 51
max_allowed_packet = 16M
tmp_table_size = 256M
max_heap_table_size = 96M
sort_buffer_size = 4M
read_buffer_size = 4M #
join_buffer_size = 4M
read_rnd_buffer_size = 64M #
bulk_insert_buffer_size = 64M #
log_error = /data/mysql3306/logs/mysql-error.log
log_timestamps = system
slow_query_log = 1
slow_query_log_file = /data/mysql3306/logs/mysql-slow.log
long_query_time = 1
log_queries_not_using_indexes = 1
log_throttle_queries_not_using_indexes =10
log_slow_admin_statements = 1
log_slow_slave_statements = 1
min_examined_row_limit = 100
log_queries_not_using_indexes = 1
log_output = FILE
relay_log = /data/mysql3306/mysql3306/mysql-relay
#binlog
server_id = 1783306
log_bin = /data/mysql3306/mysql3306/mysql-bin
binlog_expire_logs_seconds =604800
binlog_format = row
max_binlog_size = 1024M
max_binlog_cache_size = 8G
binlog_cache_size = 4M
sync_binlog = 1
master_info_repository = TABLE
relay_log_info_repository = TABLE
relay_log_recovery = ON
log_slave_updates       = 1
#innodb
default_storage_engine = InnoDB
innodb_page_size = 16384
innodb_data_home_dir = /data/mysql3306/mysql3306/
innodb_data_file_path = ibdata1:512M:autoextend
innodb_log_group_home_dir = /data/mysql3306/mysql3306
innodb_buffer_pool_instances = 1
innodb_buffer_pool_size = 1024M
innodb_buffer_pool_dump_pct = 40
innodb_page_cleaners = 8
innodb_log_file_size = 256M
innodb_log_files_in_group = 2
innodb_log_buffer_size = 32M #default 16M
innodb_flush_log_at_trx_commit = 1
innodb_lock_wait_timeout = 30
innodb_strict_mode = 1
innodb_print_all_deadlocks = 1
innodb_buffer_pool_load_at_startup = 1
innodb_buffer_pool_dump_at_shutdown = 1
innodb_buffer_pool_dump_pct = 40
innodb_open_files = 65536
innodb_file_per_table = 1
innodb_lock_wait_timeout = 30
innodb_read_io_threads = 8
innodb_write_io_threads = 8
innodb_io_capacity = 200
innodb_flush_log_at_trx_commit = 1
innodb_flush_method = O_DIRECT
innodb_purge_threads = 4
innodb_max_dirty_pages_pct = 75
transaction_isolation = READ-COMMITTED
innodb_page_cleaners = 16
explicit_defaults_for_timestamp = 1
#undo
innodb_undo_directory = /data/mysql3306/mysql3306/
innodb_undo_tablespaces = 3
innodb_undo_log_truncate = 1
innodb_max_undo_log_size = 1000M
innodb_purge_rseg_truncate_frequency = 128
#rep
slave-parallel-type=LOGICAL_CLOCK
slave-parallel-workers=16
master_info_repository=TABLE
relay_log_info_repository=TABLE
relay_log_recovery=ON
binlog_checksum=NONE
slave_preserve_commit_order=1
#group_replication_flow_control_mode=DISABLED
log_slave_updates=1
#GR
enforce_gtid_consistency=1
gtid_mode = on
transaction_write_set_extraction =XXHASH64 # off
loose-group_replication_group_name ="a876d35e-9110-11e6-a365-882b2b5s09d6" #注意格式，可以在某个库里使用SELECT UUID()生成。
loose-group_replication_local_address ="192.168.1.78:24900" #本机IP
loose-group_replication_group_seeds ="192.168.1.78:24900,192.168.1.79:24900,192.168.1.80:24900"
loose-group_replication_bootstrap_group =off # off
loose-group_replication_start_on_boot =off #非常重要
[mysqldump]
max_allowed_packet = 16M

[mysqldump]
max_allowed_packet = 16M
-----------------------------------------------------------------------
初始化
/usr/local/mysql/bin/mysqld --defaults-file=/etc/my.cnf --initialize-insecure --basedir=/usr/local/mysql --datadir=/data/mysql3306/mysql3306/ --explicit_defaults_for_timestamp --user=mysql
成功后datadir下应有22个文件。
启动
/usr/local/mysql/bin/mysqld_safe --defaults-file=/etc/my.cnf --ledir=/usr/local/mysql/bin --datadir=/data/mysql3306/mysql3306 --user=mysql &

开机自启
echo "/usr/local/mysql8/bin/mysqld_safe --defaults-file=/etc/my.cnf --ledir=/usr/local/mysql8/bin --datadir=/data/mysql3306/mysql3306 --user=mysql &">>/etc/rc.local
登录建用户及授权：
mysql
alter user root@localhost identified by '123456';
create user root@'192.168.1.%' identified by '123456';
grant all on *.* to root@'192.168.1.%';
CREATE USER rep@'%' IDENTIFIED BY '123456';
GRANT REPLICATION SLAVE ON *.* TO rep@'%';
flush privileges;

保存快照，然后复制两台虚拟机，做以下几处修改：
（1）修改虚拟机IP分别为192.168.1.79和192.168.1.80，并修改各自的主机名hostname与hosts文件对应；
（2）然后修改各自的my.cnf里的server_id和loose-group_replication_local_address改为本机IP；
（3）进入/data/mysql3306/mysql3306里删除auto.cnf文件。
然后各自都保存个快照。

二、多主搭建
1、所有节点上安装组件：
install plugin group_replication soname 'group_replication.so';
show plugins;

2、在192.168.1.78上：（只需要在这个上执行，跟单主模式相同）
CHANGE MASTER TO MASTER_USER='rep',MASTER_PASSWORD='rep' FOR CHANNEL 'group_replication_recovery';
set persist group_replication_single_primary_mode=off; --单主的都是ON
临时开启bootstrap：
SET GLOBAL group_replication_bootstrap_group=ON;
start group_replication;
SET GLOBAL group_replication_bootstrap_group=OFF;
select * from performance_schema.replication_group_members;

3、等上述操作完成之后，在192.168.1.79和192.168.1.80上分别执行：
CHANGE MASTER TO MASTER_USER='rep',MASTER_PASSWORD='123456' FOR CHANNEL 'group_replication_recovery';
set persist group_replication_single_primary_mode=off;
START GROUP_REPLICATION;

4、查看多主配置状况，在任意节点执行：select * from performance_schema.replication_group_members;

可以看到三台都在线、各自主机名正确，并且都是primary则表示MGR多主配置成功。
select * from performance_schema.replication_connection_status\G，查看是否有报错；
例如：select * from performance_schema.replication_connection_status\G;
*************************** 1. row ***************************
CHANNEL_NAME: group_replication_recovery
GROUP_NAME:
SOURCE_UUID:
THREAD_ID: NULL
SERVICE_STATE: OFF
COUNT_RECEIVED_HEARTBEATS: 0
LAST_HEARTBEAT_TIMESTAMP: 0000-00-00 00:00:00.000000
RECEIVED_TRANSACTION_SET:
LAST_ERROR_NUMBER: 0
LAST_ERROR_MESSAGE:
LAST_ERROR_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_QUEUED_TRANSACTION:
LAST_QUEUED_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_QUEUED_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_QUEUED_TRANSACTION_START_QUEUE_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_QUEUED_TRANSACTION_END_QUEUE_TIMESTAMP: 0000-00-00 00:00:00.000000
QUEUEING_TRANSACTION:
QUEUEING_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
QUEUEING_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
QUEUEING_TRANSACTION_START_QUEUE_TIMESTAMP: 0000-00-00 00:00:00.000000
*************************** 2. row ***************************
CHANNEL_NAME: group_replication_applier
GROUP_NAME: 8ef01324-97ce-11e8-b4cc-000c295ce53e
SOURCE_UUID: 8ef01324-97ce-11e8-b4cc-000c295ce53e
THREAD_ID: NULL
SERVICE_STATE: ON
COUNT_RECEIVED_HEARTBEATS: 0
LAST_HEARTBEAT_TIMESTAMP: 0000-00-00 00:00:00.000000
RECEIVED_TRANSACTION_SET: 8ef01324-97ce-11e8-b4cc-000c295ce53e:1-7:1000006:2000006
LAST_ERROR_NUMBER: 0
LAST_ERROR_MESSAGE:
LAST_ERROR_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_QUEUED_TRANSACTION: 8ef01324-97ce-11e8-b4cc-000c295ce53e:2000006
LAST_QUEUED_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 2018-08-08 09:30:19.410114
LAST_QUEUED_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_QUEUED_TRANSACTION_START_QUEUE_TIMESTAMP: 2018-08-08 09:30:19.412069
LAST_QUEUED_TRANSACTION_END_QUEUE_TIMESTAMP: 2018-08-08 09:30:19.412081
QUEUEING_TRANSACTION:
QUEUEING_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
QUEUEING_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
QUEUEING_TRANSACTION_START_QUEUE_TIMESTAMP: 0000-00-00 00:00:00.000000

三台机器都是group_replication_applier的状态为ON，group_replication_recovery的状态为OFF。
注意：
（1）如果把三台都关闭了尤其是第一台关闭了的话，再重启多主MGR的时候，要再使用bootstrap引导；
（2）如果后续测试出现问题，导致不同步，重新启动加入多主MGR的时候，要先stop group_replication再start group_replication。

贰、代理配置

1、Mysql-router的测试:
配置文件：
[DEFAULT]
logging_folder=/usr/local/mysql-router/log
runtime_folder=/usr/local/mysql-router/run
data_folder=/usr/local/mysql-router/data
connect_timeout=30
read_timeout=30

[logger]
level = INFO
[routing:failover]
bind_address = 0.0.0.0
bind_port = 7001
max_connections = 1024
mode = read-write
destinations = 192.168.1.78:3306,192.168.1.79:3306,192.168.1.80:3306

[routing:balancing]
bind_address = 0.0.0.0
bind_port = 7002
max_connections = 1024
mode = read-only
destinations = 192.168.1.78:3306,192.168.1.79:3306,192.168.1.80:3306
它需要读写端和只读端分别使用不同的端口7001和7002，暂时不做进一步测试；
-----------------------------------------------------------------------
2、Atlas代理

这个代理可以做读写分离，后续还支持分表功能，分表暂不做进一步详述，只说一下读写分离与负载均衡功能。

(1)安装

下载：

wget:https://github.com/Qihoo360/Atlas/releases/download/2.2.1/Atlas-2.2.1.el6.x86_64.rpm

安装：

yum -y localinstall Atlas-2.2.1.el6.x86_64.rpm

（2）先编辑启动文件：
vim /etc/init.d/atlas
#!/bin/bash
start()
{
/usr/local/mysql-proxy/bin/mysql-proxyd mgrmul start
}
stop()
{
/usr/local/mysql-proxy/bin/mysql-proxyd mgrmul stop
}
status()
{
/usr/local/mysql-proxy/bin/mysql-proxyd mgrmul status
}
restart()
{
/usr/local/mysql-proxy/bin/mysql-proxyd mgrmul restart
}
ATLAS="/usr/local/mysql-proxy/bin/mysql-proxyd"
[ -f $ATLAS ] || exit 1
# See how we were called.
case "$1" in
start)
start
;;
stop)
stop
;;
restart)
restart
;;
status)
status
;;
# stop sleep 3 start ;;
*)
echo $"Usage: $0 {start|stop|status|restart}"
exit 1
esac
exit 0
----------------------------------------------
（3）编辑配置文件：
在数据库里准备两个用户：
test@'192.168.1.%'，密码test，可以只赋予部分权限
admin@'%'，密码123456，可以赋予完整权限.

对使用数据库的用户的密码进行加密：

执行encrypt test获得一个密码串A1YaJzZgEjM=，执行encrypt 123456获得一个密码串/iZxz+0GRoA=；

vim /usr/local/mysql-proxy/conf/mgrmul.cnf
[mysql-proxy]
admin-username =xxx
admin-password =xxx
proxy-backend-addresses=192.168.1.78:3306,192.168.1.79:3306,192.168.1.80:3306
proxy-read-only-backend-addresses= 192.168.1.79:3306@1,192.168.1.80:3306@1 #@1代表权重
pwds =test:A1YaJzZgEjM=,admin:/iZxz+0GRoA=
daemon = true
keepalive = true
event-threads =4
log-level =warning
log-path =/usr/local/mysql-proxy/logs
instance =mgrmul
proxy-address =0.0.0.0:8060
admin-address =0.0.0.0:2345
#tables =testdb.user.id.3
client-ips =127.0.0.1, 192.168.1
---------------------------------------------------------------
（3）测试：
使用代理登录；
mysql -uadmin -p123456 -h 192.168.1.99 -P 8060
读写分离测试：Atlas代理的机制确定了它只会向配置文件中记录的可写节点的第一个可用的去写，如果第一个不可用就会去访问第二个，直到找到一个可用的并一直只向该主机去写；读的话则从配置文件中的所有可读节点去读；

负载均衡测试：select @@hostname，可以查看主机名来自不同的可读主机；

failover测试：
pkill 192.168.1.78上的mysql;
继续使用代理登录，仍可以实现读写操作。