MySQL8.0的MGR多主搭建与测试
本文部分内容有参考其他文章。
环境：
虚拟机VMware，
OS：CentOS Linux release 7.3.1611 (Core)
按照下述的单机Mysql8.0的安装方式进行安装完之后，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/etc/my.cnf的serverid，还有要去数据路径下删掉auto.cnf不然配置不成。
vim /etc/hosts
192.168.1.78 mgr-mul1
192.168.1.79 mgr-mul2
192.168.1.80 mgr-mul3
要用hostname修改每台机器的主机名确保不相同，不然配置不成功。
一、单机安装(192.168.1.78)
官网下载8.0.11版本的tar.gz包解压，复制更名到/usr/local/mysql
创建数据路径及日志路径并把他们权限改给mysql用户：
/data/mysql3306/mysql3306
/data/mysql3306/logs
配置文件如下：注意修改server-id及组相关ip
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 #??IO
join_buffer_size = 4M
read_rnd_buffer_size = 64M #??篓虏
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" #可在库里使用SELECT UUID()生成，一个组的所有机器都相同。
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
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 &
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：
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 'rep';
GRANT REPLICATION SLAVE ON *.* TO rep@'%';
flush privileges;
保存快照（before MGR-Multi），然后复制两台，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。
然后各自保存快照（before MGR-Multi）；
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依次执行：
CHANGE MASTER TO MASTER_USER='rep',MASTER_PASSWORD='rep' FOR CHANNEL 'group_replication_recovery';
set persist group_replication_single_primary_mode=off;
START GROUP_REPLICATION;
4、验证：
任意库执行：select * from performance_schema.replication_group_members;
select * from performance_schema.replication_group_members;
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME | MEMBER_ID | MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 5651f374-97c8-11e8-9966-000c295ce53e | mgr-mul1 | 3306 | ONLINE | PRIMARY | 8.0.11 |
| group_replication_applier | 749eb449-97e1-11e8-ab87-000c29ee3a4b | mgr-mul3 | 3306 | ONLINE | PRIMARY | 8.0.11 |
| group_replication_applier | 7753edfa-97e1-11e8-a10c-000c296952a6 | mgr-mul2 | 3306 | ONLINE | PRIMARY | 8.0.11 |
+---------------------------+--------------------------------------+-------------+-------------+--------------+-------------+----------------+
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。
5、注意：#############################################################################
如果把三台全都关闭了再重启多主MGR的时候，要再使用bootstrap；
#############################################################################
三、测试
任意库建库建表，其他库均可同步。
四、问题
感觉MGR多主似乎不太稳定，经常会出现某个库掉线的情况又不会自动拉起来；用循环脚本测试一旦出现较高的并发，就会掉节点，不论是否用代理测试都掉节点，好像不太实用，也可能是因为我配置的什么地方参数有问题，希望有高人给指点一下。
拉起来节点只需要依次执行stop group_replication和start group_replication即可。