实验场景描述及目的：
GTID的主从模式，目前是一主一从，希望再加一个slave做成一主两从的模式。由于测试的数据库不大，通过mysqldump备份数据，如果数据库比较大的情况下，可以想办法通过其它slave的备份来搭建新的slave。
1.备份master
mysql> FLUSH TABLE WITH READ LOCK; Query OK, 0 rows affected (0.01 sec)
[root@qht131 backup]# mysqldump -u root -p --lock-all-tables --master-data=2 --flush-logs --all-databases --triggers --routines --events > full.sql
[root@qht131 backup]# scp full.sql 172.17.61.133:/u01/backup
记录当前的gtid:
mysql> show global variables like 'gtid_%'; +----------------------------------+-------------------------------------------+ | Variable_name | Value | +----------------------------------+-------------------------------------------+ | gtid_executed | 8d8746fb-2cc6-11e8-b1b6-000c295c63e0:1-11 | | gtid_executed_compression_period | 1000 | | gtid_mode | ON | | gtid_owned | | | gtid_purged | | +----------------------------------+-------------------------------------------+ 5 rows in set (0.09 sec)
如果是Xtrabackup的话，Xtrabackup_binlog_info文件中包含global.gtid_purged='XXXXXX:XXXX'的信息。
2.在新的slave上建好的数据库并恢复Master的备份
2.1.建库方法:https://blog.csdn.net/jolly10/article/details/79566640
2.2.恢复全备：
[root@qht133 backup]# mysql -uroot -p < full.sql
2.3新slave上修改/etc/my.cnf，开启gtid
[root@qht133 backup]# cat /etc/my.cnf [mysqld] socket = /usr/local/mysql/mysql.sock character_set_server= utf8 init_connect= 'SET NAMES utf8' basedir= /usr/local/mysql datadir= /u01/mysql socket = /u01/mysql/mysql.sock log-error= /u01/log/mysql/mysql_3306.err pid-file= /u01/mysql/mysqld.pid lower_case_table_names = 1 sql_mode= STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION secure-file-priv = /u01/backup gtid_mode=on #开启gtid enforce_gtid_consistency=on #强制gtid一致性 server-id=10003 #server id skip_slave_start=1 #开启备库后不自动应用恢复进程 #log_bin = /u01/mysql/mysql_bin #skip-grant-tables #innodb_flush_log_at_trx_commit=1 #sync_binlog=1 relay-log=/u01/mysql/slave_relay_bin expire_logs_days=10 relay_log_recovery=on relay_log_info_repository=TABLE
重启数据库：
[root@qht133 backup]# service mysql restart Shutting down MySQL.. [ OK ] Starting MySQL.. [ OK ]
3.为了检验新的slave复制的正确性，在master上做些操作，看slave能否正确的同步。
mysql> use l5m Database changed mysql> select count(*) from test_emp; +----------+ | count(*) | +----------+ | 1099000 | +----------+ 1 row in set (0.67 sec) mysql> delete from test_emp limit 1000; Query OK, 1000 rows affected (0.08 sec) mysql> select count(*) from test_emp; +----------+ | count(*) | +----------+ | 1098000 | +----------+ 1 row in set (0.18 sec) mysql> commit; Query OK, 0 rows affected (0.00 sec) mysql> show global variables like 'gtid_%'; +----------------------------------+-------------------------------------------+ | Variable_name | Value | +----------------------------------+-------------------------------------------+ | gtid_executed | 8d8746fb-2cc6-11e8-b1b6-000c295c63e0:1-12 | | gtid_executed_compression_period | 1000 | | gtid_mode | ON | | gtid_owned | | | gtid_purged | | +----------------------------------+-------------------------------------------+ 5 rows in set (0.01 sec)
4.在新的slave上开启复制：
先检查一下当前的gtid：
mysql> show global variables like 'gtid_%'; +----------------------------------+-------------------------------------------+ | Variable_name | Value | +----------------------------------+-------------------------------------------+ | gtid_executed | 8d8746fb-2cc6-11e8-b1b6-000c295c63e0:1-11 | | gtid_executed_compression_period | 1000 | | gtid_mode | ON | | gtid_owned | | | gtid_purged | 8d8746fb-2cc6-11e8-b1b6-000c295c63e0:1-11 | +----------------------------------+-------------------------------------------+ 5 rows in set (0.00 sec)
新的slave库是从master恢复过来的，gtid_purged现在自动就有了值1-11，并不需要手动的执行reset master; set global gtid_purged = 'xxxxx'，看来mysql5.7这方面又加强了！（只有在@@global.gtid_executed为空的情况下，才可以动态设置@@global.gtid_purged。因此可以通过RESET MASTER的方式来清空@@global.gtid_executed。）
下面直接开启复制就行了：
mysql> change master to
-> master_host='172.17.61.131',
-> master_port=3306,
-> master_user='repl',
-> master_password='repl',
-> master_auto_position=1;
Query OK, 0 rows affected, 2 warnings (0.07 sec)
mysql> start slave;
mysql> show global variables like 'gtid_%'; +----------------------------------+-------------------------------------------+ | Variable_name | Value | +----------------------------------+-------------------------------------------+ | gtid_executed | 8d8746fb-2cc6-11e8-b1b6-000c295c63e0:1-12 | | gtid_executed_compression_period | 1000 | | gtid_mode | ON | | gtid_owned | | | gtid_purged | 8d8746fb-2cc6-11e8-b1b6-000c295c63e0:1-12 | +----------------------------------+-------------------------------------------+ 5 rows in set (0.01 sec)
验证一下全备之后master上修改的数据有没有同步过来：
mysql> use l5m; Database changed mysql> select count(*) from test_emp; +----------+ | count(*) | +----------+ | 1098000 | +----------+ 1 row in set (0.40 sec)
没有问题，数据都同步到了新的slave。
参考：
https://www.cnblogs.com/zejin2008/p/7705473.html
http://www.cnblogs.com/luckcs/articles/6295992.html