要讨论如何恢复从库，我们得先来了解如下一些概念：
GTID_EXECUTED:它是一组包含已经记录在二进制日志文件中的事务集合
GTID_PURGED:它是一组包含已经从二进制日志删除掉的事务集合。
在继续讨论时，我们先来看下如何新建一个基于GTID的slave。
通过了解上面的两个参数，我们现在只需要：
1.从主库上做一个备份时记录备份时gtid_executed的值。
2.在新的slave上恢复此备份时设置从库的gtid_purged的值为备份时master上gtid_executed的值。
通过mysqldump可以完成我们需要的功能。
目前主库上的状态（3301）：
[zejin] 3301>show global variables like 'gtid_executed';
+---------------+-------------------------------------------+
| Variable_name | Value |
+---------------+-------------------------------------------+
| gtid_executed | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-15 |
+---------------+-------------------------------------------+
1 row in set (0.00 sec)
[zejin] 3301>show global variables like 'gtid_purged';
+---------------+-------------------------------------------+
| Variable_name | Value |
+---------------+-------------------------------------------+
| gtid_purged | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-13 |
+---------------+-------------------------------------------+
1 row in set (0.00 sec)
step1:用mysqldump做一个全备
mysqldump --all-databases --single-transaction --triggers --routines --events --host=127.0.0.1 --port=3301 --user=root --password=123 > dump3301.sql
打开dump3301.sql我们可以看到如下语句：
SET @@GLOBAL.GTID_PURGED='a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-15';
此值即为master3301上gtid_executed的值。
step2:全新启动一个新的库3303,注意在配置文件中配置enforce_gtid_consistency及gtid_mode=on
mysqld_safe --defaults-file=/home/mysql/my3303.cnf &
此时新库3303上的状态应该是这样的：
[(none)] 3303>show global variables like 'gtid_executed';
+---------------+-------+
| Variable_name | Value |
+---------------+-------+
| gtid_executed | |
+---------------+-------+
1 row in set (0.01 sec)
[(none)] 3303>show global variables like 'gtid_purged';
+---------------+-------+
| Variable_name | Value |
+---------------+-------+
| gtid_purged | |
+---------------+-------+
1 row in set (0.00 sec)
step3:导入备份文件并查看状态值：
mysql -uroot -h127.0.0.1 -p123 -P3303 < dump3301.sql
[(none)] 3303>show global variables like 'gtid_executed';
+---------------+-------------------------------------------+
| Variable_name | Value |
+---------------+-------------------------------------------+
| gtid_executed | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-15 |
+---------------+-------------------------------------------+
1 row in set (0.02 sec)
[(none)] 3303>show global variables like 'gtid_purged';
+---------------+-------------------------------------------+
| Variable_name | Value |
+---------------+-------------------------------------------+
| gtid_purged | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-15 |
+---------------+-------------------------------------------+
1 row in set (0.00 sec)
step4:做主从change语句
[zejin] 3303>change master to master_host='192.168.1.240',master_port=3301,master_user='repl',master_password='123',master_auto_position=1;
Query OK, 0 rows affected, 2 warnings (0.01 sec)
[zejin] 3303>start slave;
Query OK, 0 rows affected (0.00 sec)
[zejin] 3303>show slave status\G
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.1.240
Master_User: repl
Master_Port: 3301
Connect_Retry: 60
Master_Log_File: binlog57.000014
Read_Master_Log_Pos: 194
Relay_Log_File: zejin240-relay-bin.000002
Relay_Log_Pos: 365
Relay_Master_Log_File: binlog57.000014
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 0
Last_Error:
Skip_Counter: 0
Exec_Master_Log_Pos: 194
Relay_Log_Space: 575
Until_Condition: None
Until_Log_File:
Until_Log_Pos: 0
Master_SSL_Allowed: No
Master_SSL_CA_File:
Master_SSL_CA_Path:
Master_SSL_Cert:
Master_SSL_Cipher:
Master_SSL_Key:
Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 0
Last_IO_Error:
Last_SQL_Errno: 0
Last_SQL_Error:
Replicate_Ignore_Server_Ids:
Master_Server_Id: 3301
Master_UUID: a97983fc-5a29-11e6-9d28-000c29d4dc3f
Master_Info_File: /home/mysql/I3303/master.info
SQL_Delay: 0
SQL_Remaining_Delay: NULL
Slave_SQL_Running_State: Slave has read all relay log; waiting for more updates
Master_Retry_Count: 86400
Master_Bind:
Last_IO_Error_Timestamp:
Last_SQL_Error_Timestamp:
Master_SSL_Crl:
Master_SSL_Crlpath:
Retrieved_Gtid_Set:
Executed_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-15
Auto_Position: 1
Replicate_Rewrite_DB:
Channel_Name:
Master_TLS_Version:
1 row in set (0.00 sec)
至此完成了加入一台新的slave的GTID主从环境。
假如我们目前拥有一主带两从的环境：
master（3301）
slave（3302）
slave（3303）
我们来考虑这么一种异常情况，由于种种原因，有可能主库上已经purge掉了一些binlog，但从库都还没有接收到（如slave停了一段时间，而master已经把一些binlog给purge掉了。）
主库目前的状态是：
[zejin] 3301>show global variables like 'gtid_executed';
+---------------+-------------------------------------------+
| Variable_name | Value |
+---------------+-------------------------------------------+
| gtid_executed | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-21 |
+---------------+-------------------------------------------+
1 row in set (0.00 sec)
[zejin] 3301>show global variables like 'gtid_purged';
+---------------+-------------------------------------------+
| Variable_name | Value |
+---------------+-------------------------------------------+
| gtid_purged | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-20 |
+---------------+-------------------------------------------+
1 row in set (0.00 sec)
[zejin] 3301>select * from t_users;
+----+------+
| id | name |
+----+------+
| 1 | chen |
| 2 | ok |
| 3 | li |
+----+------+
3 rows in set (0.00 sec)
在从库3303上，我们可以看到如下错误提示：
Last_IO_Error: Got fatal error 1236 from master when reading data from binary log: 'The slave is connecting using CHANGE MASTER TO MASTER_AUTO_POSITION = 1, but the master has purged binary logs containing GTIDs that the slave requires.'
[zejin] 3303>show slave status\G
*************************** 1. row ***************************
Slave_IO_State:
Master_Host: 192.168.1.240
Master_User: repl
Master_Port: 3301
Connect_Retry: 60
Master_Log_File: binlog57.000014
Read_Master_Log_Pos: 457
Relay_Log_File: zejin240-relay-bin.000003
Relay_Log_Pos: 4
Relay_Master_Log_File: binlog57.000014
Slave_IO_Running: No
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 0
Last_Error:
Skip_Counter: 0
Exec_Master_Log_Pos: 457
Relay_Log_Space: 194
Until_Condition: None
Until_Log_File:
Until_Log_Pos: 0
Master_SSL_Allowed: No
Master_SSL_CA_File:
Master_SSL_CA_Path:
Master_SSL_Cert:
Master_SSL_Cipher:
Master_SSL_Key:
Seconds_Behind_Master: NULL
Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 1236
Last_IO_Error: Got fatal error 1236 from master when reading data from binary log: 'The slave is connecting using CHANGE MASTER TO MASTER_AUTO_POSITION = 1, but the master has purged binary logs containing GTIDs that the slave requires.'
Last_SQL_Errno: 0
Last_SQL_Error:
Replicate_Ignore_Server_Ids:
Master_Server_Id: 3301
Master_UUID: a97983fc-5a29-11e6-9d28-000c29d4dc3f
Master_Info_File: /home/mysql/I3303/master.info
SQL_Delay: 0
SQL_Remaining_Delay: NULL
Slave_SQL_Running_State: Slave has read all relay log; waiting for more updates
Master_Retry_Count: 86400
Master_Bind:
Last_IO_Error_Timestamp: 160809 17:25:39
Last_SQL_Error_Timestamp:
Master_SSL_Crl:
Master_SSL_Crlpath:
Retrieved_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:16
Executed_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-16
Auto_Position: 1
Replicate_Rewrite_DB:
Channel_Name:
Master_TLS_Version:
1 row in set (0.00 sec)
[zejin] 3303>select * from t_users;
+----+------+
| id | name |
+----+------+
| 1 | li |
| 2 | zhou |
+----+------+
2 rows in set (0.00 sec)
主从已经中断，数据也已不一致。
接下来我们来看如何恢复：
由于GTID具有全局唯一性，那么其它正常的gtid已经被复制到了slave3302上，所以我们可以把3303指向3302，同步完毕后再指回master3301（此前提基于3302的binlog还没被purge掉，即存在3303没有从master3301接收到的GTID事务）
操作方法如下：
[zejin] 3303>change master to master_host='192.168.1.240',master_port=3302,master_user='repl',master_password='123',master_auto_position=1;
[zejin] 3303>start slave;
Query OK, 0 rows affected (0.03 sec)
[zejin] 3303>show slave status\G
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.1.240
Master_User: repl
Master_Port: 3302
Connect_Retry: 60
Master_Log_File: binlog57.000007
Read_Master_Log_Pos: 1723
Relay_Log_File: zejin240-relay-bin.000002
Relay_Log_Pos: 1687
Relay_Master_Log_File: binlog57.000007
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 0
Last_Error:
Skip_Counter: 0
Exec_Master_Log_Pos: 1723
Relay_Log_Space: 1937
Until_Condition: None
Until_Log_File:
Until_Log_Pos: 0
Master_SSL_Allowed: No
Master_SSL_CA_File:
Master_SSL_CA_Path:
Master_SSL_Cert:
Master_SSL_Cipher:
Master_SSL_Key:
Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 0
Last_IO_Error:
Last_SQL_Errno: 0
Last_SQL_Error:
Replicate_Ignore_Server_Ids:
Master_Server_Id: 3302
Master_UUID: 5cee6f9f-5ab8-11e6-a081-000c29d4dc3f
Master_Info_File: /home/mysql/I3303/master.info
SQL_Delay: 0
SQL_Remaining_Delay: NULL
Slave_SQL_Running_State: Slave has read all relay log; waiting for more updates
Master_Retry_Count: 86400
Master_Bind:
Last_IO_Error_Timestamp:
Last_SQL_Error_Timestamp:
Master_SSL_Crl:
Master_SSL_Crlpath:
Retrieved_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:17-21
Executed_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-21
Auto_Position: 1
Replicate_Rewrite_DB:
Channel_Name:
Master_TLS_Version:
1 row in set (0.00 sec)
[zejin] 3303>select * from t_users;
+----+------+
| id | name |
+----+------+
| 1 | chen |
| 2 | ok |
| 3 | li |
+----+------+
3 rows in set (0.00 sec)
数据也已经完全与主的一致了，复制正常后再change到3301master上。
[zejin] 3303>change master to master_host='192.168.1.240',master_port=3301,master_user='repl',master_password='123',master_auto_position=1;
Query OK, 0 rows affected, 2 warnings (0.01 sec)
[zejin] 3303>start slave;
Query OK, 0 rows affected (0.00 sec)
上面这种情况是基于还有另一个从库已经接收到了master的所有binlog的情况下，那如果结果只是M-S，也发生了如上的问题，那又该如何恢复，我们有如下两种方法：
目前Master上状态为：
[zejin] 3301>show global variables like '%gtid%';
+----------------------------------+-------------------------------------------+
| Variable_name | Value |
+----------------------------------+-------------------------------------------+
| gtid_executed | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-27 |
……
| gtid_purged | a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-25 |
……
+----------------------------------+-------------------------------------------+
8 rows in set (0.00 sec)
Slave上状态为：
[zejin] 3303>show slave status \G
*************************** 1. row ***************************
Slave_IO_State:
Master_Host: 192.168.1.240
Master_User: repl
Master_Port: 3301
Connect_Retry: 60
Master_Log_File: binlog57.000016
Read_Master_Log_Pos: 729
Relay_Log_File: zejin240-relay-bin.000003
Relay_Log_Pos: 4
Relay_Master_Log_File: binlog57.000016
Slave_IO_Running: No
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 0
Last_Error:
Skip_Counter: 0
Exec_Master_Log_Pos: 729
Relay_Log_Space: 194
Until_Condition: None
Until_Log_File:
Until_Log_Pos: 0
Master_SSL_Allowed: No
Master_SSL_CA_File:
Master_SSL_CA_Path:
Master_SSL_Cert:
Master_SSL_Cipher:
Master_SSL_Key:
Seconds_Behind_Master: NULL
Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 1236
Last_IO_Error: Got fatal error 1236 from master when reading data from binary log: 'The slave is connecting using CHANGE MASTER TO MASTER_AUTO_POSITION = 1, but the master has purged binary logs containing GTIDs that the slave requires.'
Last_SQL_Errno: 0
Last_SQL_Error:
Replicate_Ignore_Server_Ids:
Master_Server_Id: 3301
Master_UUID: a97983fc-5a29-11e6-9d28-000c29d4dc3f
Master_Info_File: /home/mysql/I3303/master.info
SQL_Delay: 0
SQL_Remaining_Delay: NULL
Slave_SQL_Running_State: Slave has read all relay log; waiting for more updates
Master_Retry_Count: 86400
Master_Bind:
Last_IO_Error_Timestamp: 160809 17:54:42
Last_SQL_Error_Timestamp:
Master_SSL_Crl:
Master_SSL_Crlpath:
Retrieved_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:22
Executed_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-22
Auto_Position: 1
Replicate_Rewrite_DB:
Channel_Name:
Master_TLS_Version:
1 row in set (0.00 sec)
和之前同样类型的错误，我们恢复的思路为：
把slave上的gtid_purged设置为master还没有被purge掉的值，最后借助第三方一致性同步工具来做数据的一致性同步。
我们需要先在slave上做一下reset master来清除gtid的一些信息，直接设置会报如下错误：
[zejin] 3303>set global GTID_PURGED="a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-26";
ERROR 1840 (HY000): @@GLOBAL.GTID_PURGED can only be set when @@GLOBAL.GTID_EXECUTED is empty.
正确操作步骤如下（在slave上执行）：
[zejin] 3303>reset master;
Query OK, 0 rows affected (0.02 sec)
[zejin] 3303>set global GTID_PURGED="a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-26";
Query OK, 0 rows affected (0.00 sec)
[zejin] 3303>start slave;
Query OK, 0 rows affected (0.00 sec)
[zejin] 3303>show slave status \G
*************************** 1. row ***************************
Slave_IO_State: Waiting for master to send event
Master_Host: 192.168.1.240
Master_User: repl
Master_Port: 3301
Connect_Retry: 60
Master_Log_File: binlog57.000018
Read_Master_Log_Pos: 728
Relay_Log_File: zejin240-relay-bin.000004
Relay_Log_Pos: 718
Relay_Master_Log_File: binlog57.000018
Slave_IO_Running: Yes
Slave_SQL_Running: Yes
Replicate_Do_DB:
Replicate_Ignore_DB:
Replicate_Do_Table:
Replicate_Ignore_Table:
Replicate_Wild_Do_Table:
Replicate_Wild_Ignore_Table:
Last_Errno: 0
Last_Error:
Skip_Counter: 0
Exec_Master_Log_Pos: 728
Relay_Log_Space: 968
Until_Condition: None
Until_Log_File:
Until_Log_Pos: 0
Master_SSL_Allowed: No
Master_SSL_CA_File:
Master_SSL_CA_Path:
Master_SSL_Cert:
Master_SSL_Cipher:
Master_SSL_Key:
Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
Last_IO_Errno: 0
Last_IO_Error:
Last_SQL_Errno: 0
Last_SQL_Error:
Replicate_Ignore_Server_Ids:
Master_Server_Id: 3301
Master_UUID: a97983fc-5a29-11e6-9d28-000c29d4dc3f
Master_Info_File: /home/mysql/I3303/master.info
SQL_Delay: 0
SQL_Remaining_Delay: NULL
Slave_SQL_Running_State: Slave has read all relay log; waiting for more updates
Master_Retry_Count: 86400
Master_Bind:
Last_IO_Error_Timestamp:
Last_SQL_Error_Timestamp:
Master_SSL_Crl:
Master_SSL_Crlpath:
Retrieved_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:22:27
Executed_Gtid_Set: a97983fc-5a29-11e6-9d28-000c29d4dc3f:1-27
Auto_Position: 1
Replicate_Rewrite_DB:
Channel_Name:
Master_TLS_Version:
1 row in set (0.00 sec)
当然执行完这个之后数据是不一致的，那么此时就可以通过pt-table-checksum和pt-table-sync来做数据的一致性恢复了。
我们还有另一种方法，那就是重建slave，方法如本文最开始的那样新建一个slave，但是在由于目前slave上已经有gtid的一些信息，所以在恢复时得先在slave上reset master，具体操作如下：
在slave上操作：
reset master
source dump3301.sql;
change master to master_host='192.168.1.240',master_port=3301,master_user='repl',master_password='123',master_auto_position=1;
start slave;
show slave status\G
至此完成slave同步异常的恢复。
