MySQL备份工具，支持各种参数选项，使用不同的选项极有可能影响备份处理过程。本文使用我们常规认为合理的备份参数，测试/验证是否存在容易忽视的坑
# 常规备份参数 # mysqldump shell> mysqldump --single-transaction --master-data=2 -B replcrash >dbname_dump_serverid_`date +%Y%m%d`.sql # mysqlpump shell> mysqlpump --single-transaction -B replcrash >dbname_pump_serverid_`date +%Y%m%d`.sql # mydumper shell> mydumper -B replcrash -o /data/backup/mydumper # XtraBackup # backup shell> innobackupex [--defaults-file=MY.CNF] BACKUP-ROOT-DIR # apply-log shell> innobackupex --apply-log [--defaults-file=MY.CNF] BACKUP-DIR # copy-back shell> innobackupex --copy-back [--defaults-file=MY.CNF] BACKUP-DIR
| ROLE | HOSTNAME | BASEDIR | DATADIR | IP | PORT |
| Master | ZST1 | /usr/local/mysql | /data/mysql/mysql3306/data | 192.168.85.132 | 3306 |
| Slave | ZST1 | /usr/local/mysql | /data/mysql/mysql3308/data | 192.168.85.132 | 3308 |
官方社区版MySQL 5.7.19 基于Row+Position搭建的一主一从异步复制结构：Master->{Slave}。每次使用备份文件还原数据库后，重新搭建这个复制结构
备份工具版本：mysqldump、mysqlpump是MySQL 5.7.19中自带的；mydumper version 0.9.3、innobackupex version 2.4.8
一、mysqldump
1.1、DML操作对备份的影响
创建两张测试表
# 创建两张测试表(192.168.85.132,3306) use replcrash; create table py_user_innodb( uid int not null auto_increment, name varchar(32), add_time datetime default current_timestamp, server_id varchar(10), primary key(uid), key(name) )engine=innodb; create table py_user_myisam( uid int not null auto_increment, name varchar(32), add_time datetime default current_timestamp, server_id varchar(10), primary key(uid), key(name) )engine=myisam;
运行下面的脚本持续往测试表中写入数据
#!/user/bin/python import string import random import MySQLdb import time conn = MySQLdb.connect(host=‘192.168.85.132‘, port=3306, user=‘mydba‘, passwd=‘mysql5719‘, db=‘replcrash‘) """ create table py_user( uid int not null auto_increment, name varchar(32), add_time datetime default current_timestamp, server_id varchar(10), primary key(uid), key(name) ); """ while True: r_name = ‘‘.join(random.choice(string.ascii_uppercase + string.digits) for _ in range(random.randint(20,30))) print r_name cursor = conn.cursor() cursor.execute("insert into py_user_myisam(name,add_time,server_id) values(‘%s‘,now(),@@server_id);" % str(r_name)) cursor.execute("insert into py_user_innodb(name,add_time,server_id) values(‘%s‘,now(),@@server_id);" % str(r_name)) conn.commit() time.sleep(0.001)
开启general_log，用来查看mysqldump执行过程
# 开启general_log mydba@192.168.85.132,3306 [replcrash]> set global general_log_file=‘/data/mysql/mysql3306/data/mysql-general.log‘; mydba@192.168.85.132,3306 [replcrash]> set global general_log=1; # 清空general_log [root@ZST1 logs]# cat /dev/null > /data/mysql/mysql3306/data/mysql-general.log # 备份replcrash数据库 [root@ZST1 backup]# mysqldump -h127.0.0.1 -P3306 -uroot -p --single-transaction --master-data=2 replcrash >/data/backup/replcrash_dump_1323306_`date +%Y%m%d`.sql
使用备份文件搭建复制
# 还原实例清空GTID信息 mydba@192.168.85.132,3308 [replcrash]> reset master; # 还原数据 [root@ZST1 backup]# mysql -h127.0.0.1 -P3308 -uroot -p replcrash </data/backup/replcrash_dump_1323306_`date +%Y%m%d`.sql # 搭建复制 mydba@192.168.85.132,3308 [replcrash]> change master to master_host=‘192.168.85.132‘, master_port=3306, master_user=‘repl‘, master_password=‘repl‘, master_auto_position=1; # 启动复制，查看复制状态 mydba@192.168.85.132,3308 [replcrash]> start slave; mydba@192.168.85.132,3308 [replcrash]> show slave status\G *************************** 1. row *************************** Master_Log_File: mysql-bin.000183 Read_Master_Log_Pos: 1541377 Relay_Log_File: relay-bin.000002 Relay_Log_Pos: 741 Relay_Master_Log_File: mysql-bin.000183 Slave_IO_Running: Yes Slave_SQL_Running: No Exec_Master_Log_Pos: 1042768 Last_SQL_Errno: 1062 Last_SQL_Error: Could not execute Write_rows event on table replcrash.py_user_myisam; Duplicate entry ‘332‘ for key ‘PRIMARY‘, Error_code: 1062; handler error HA_ERR_FOUND_DUPP_KEY; the event‘‘s master log mysql-bin.000183, end_log_pos 1043062 Retrieved_Gtid_Set: 8ab82362-9c37-11e7-a858-000c29c1025c:251874-253268 Executed_Gtid_Set: 8ab82362-9c37-11e7-a858-000c29c1025c:1-251874 Auto_Position: 1
从上面的结果中可以看到，主键冲突了，在从库查询一下这个表中大于等于冲突key的数据
# 查询从库出错表大于等于冲突key的数据 mydba@192.168.85.132,3308 [replcrash]> select * from replcrash.py_user_myisam where uid>=332; +-----+--------------------------------+---------------------+-----------+ | uid | name | add_time | server_id | +-----+--------------------------------+---------------------+-----------+ | 332 | X1LME9HO5V7WXNOKBVZE | 2018-01-02 09:05:07 | 1323306 | | 333 | 2PBFQ7KS4BPIJ27G88EYXWEDSX5 | 2018-01-02 09:05:07 | 1323306 | | 334 | E85Y2SS9UD0FZG4YGCNTRSWA8L | 2018-01-02 09:05:07 | 1323306 | | 335 | Y2TQOEVJ58NN7EREL4WRZ | 2018-01-02 09:05:07 | 1323306 | | 336 | O0MEATAXYIAE2V2IZG96YVQ56WEUHF | 2018-01-02 09:05:07 | 1323306 | | 337 | A6QKRWEXHRGUA3V2CH61VXUNBVA3H2 | 2018-01-02 09:05:07 | 1323306 | | 338 | NYCSI1HS61BN6QAVVYTZSC | 2018-01-02 09:05:07 | 1323306 | | 339 | 7CFC1JQPIQGNC97MDTT8ZIMIZL7D | 2018-01-02 09:05:07 | 1323306 | | 340 | GA78AR4Z12WQTEAM41JB | 2018-01-02 09:05:07 | 1323306 | +-----+--------------------------------+---------------------+-----------+ 9 rows in set (0.08 sec)
我们查看mysqldump备份文件获取的binlog pos
[root@ZST1 backup]# more replcrash_dump_1323306_20180102.sql -- GTID state at the beginning of the backup SET @@GLOBAL.GTID_PURGED=‘8ab82362-9c37-11e7-a858-000c29c1025c:1-251873‘; -- Position to start replication or point-in-time recovery from -- CHANGE MASTER TO MASTER_LOG_FILE=‘mysql-bin.000183‘, MASTER_LOG_POS=1042441; [root@ZST1 backup]#
这里的pos信息是mysqldump通过SHOW MASTER STATUS获取。查看mysqldump得到的general-log;
[root@ZST1 data]# vim /data/mysql/mysql3306/data/mysql-general.log ... 2018-01-02T01:05:07.693104Z 10 Query FLUSH /*!40101 LOCAL */ TABLES 2018-01-02T01:05:07.694738Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘7ATZSNFNIBW5DZNMNZYBMV‘,now(),@@server_id) 2018-01-02T01:05:07.701616Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘7ATZSNFNIBW5DZNMNZYBMV‘,now(),@@server_id) 2018-01-02T01:05:07.702139Z 10 Query FLUSH TABLES WITH READ LOCK 2018-01-02T01:05:07.702344Z 9 Query commit 2018-01-02T01:05:07.702411Z 10 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 2018-01-02T01:05:07.702597Z 10 Query START TRANSACTION /*!40100 WITH CONSISTENT SNAPSHOT */ 2018-01-02T01:05:07.702721Z 10 Query SHOW VARIABLES LIKE ‘gtid\_mode‘ 2018-01-02T01:05:07.713019Z 10 Query SELECT @@GLOBAL.GTID_EXECUTED 2018-01-02T01:05:07.713179Z 10 Query SHOW MASTER STATUS 2018-01-02T01:05:07.725821Z 10 Query UNLOCK TABLES 2018-01-02T01:05:07.732125Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘X1LME9HO5V7WXNOKBVZE‘,now(),@@server_id) 2018-01-02T01:05:07.733237Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘X1LME9HO5V7WXNOKBVZE‘,now(),@@server_id) 2018-01-02T01:05:07.734240Z 9 Query commit 2018-01-02T01:05:07.740508Z 10 Query SELECT LOGFILE_GROUP_NAME, FILE_NAME, TOTAL_EXTENTS, INITIAL_SIZE, ENGINE, EXTRA FROM INFORMATION_SCHEMA.FILES WHERE FILE_TYPE = ‘UNDO LOG‘ AND FILE_NAME IS NOT NULL AND LOGFILE_GROUP_NAME IS NOT NULL AND LOGFILE_GROUP_NAME IN (SELECT DISTINCT LOGFILE_GROUP_NAME FROM INFORMATION_SCHEMA.FILES WHERE FILE_TYPE = ‘DATAFILE‘ AND TABLESPACE_NAME IN (SELECT DISTINCT TABLESPACE_NAME FROM INFORMATION_SCHEMA.PARTITIONS WHERE TABLE_SCHEMA IN (‘replcrash‘))) GROUP BY LOGFILE_GROUP_NAME, FILE_NAME, ENGINE, TOTAL_EXTENTS, INITIAL_SIZE ORDER BY LOGFILE_GROUP_NAME 2018-01-02T01:05:07.741895Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘2PBFQ7KS4BPIJ27G88EYXWEDSX5‘,now(),@@server_id) 2018-01-02T01:05:07.742720Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘2PBFQ7KS4BPIJ27G88EYXWEDSX5‘,now(),@@server_id) 2018-01-02T01:05:07.743257Z 9 Query commit 2018-01-02T01:05:07.749840Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘E85Y2SS9UD0FZG4YGCNTRSWA8L‘,now(),@@server_id) 2018-01-02T01:05:07.750588Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘E85Y2SS9UD0FZG4YGCNTRSWA8L‘,now(),@@server_id) 2018-01-02T01:05:07.750989Z 9 Query commit 2018-01-02T01:05:07.754180Z 10 Query SELECT DISTINCT TABLESPACE_NAME, FILE_NAME, LOGFILE_GROUP_NAME, EXTENT_SIZE, INITIAL_SIZE, ENGINE FROM INFORMATION_SCHEMA.FILES WHERE FILE_TYPE = ‘DATAFILE‘ AND TABLESPACE_NAME IN (SELECT DISTINCT TABLESPACE_NAME FROM INFORMATION_SCHEMA.PARTITIONS WHERE TABLE_SCHEMA IN (‘replcrash‘)) ORDER BY TABLESPACE_NAME, LOGFILE_GROUP_NAME 2018-01-02T01:05:07.756229Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘Y2TQOEVJ58NN7EREL4WRZ‘,now(),@@server_id) 2018-01-02T01:05:07.757030Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘Y2TQOEVJ58NN7EREL4WRZ‘,now(),@@server_id) 2018-01-02T01:05:07.757598Z 9 Query commit 2018-01-02T01:05:07.763629Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘O0MEATAXYIAE2V2IZG96YVQ56WEUHF‘,now(),@@server_id) 2018-01-02T01:05:07.764626Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘O0MEATAXYIAE2V2IZG96YVQ56WEUHF‘,now(),@@server_id) 2018-01-02T01:05:07.765654Z 9 Query commit 2018-01-02T01:05:07.766769Z 10 Query SHOW VARIABLES LIKE ‘ndbinfo\_version‘ 2018-01-02T01:05:07.773997Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘A6QKRWEXHRGUA3V2CH61VXUNBVA3H2‘,now(),@@server_id) 2018-01-02T01:05:07.774757Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘A6QKRWEXHRGUA3V2CH61VXUNBVA3H2‘,now(),@@server_id) 2018-01-02T01:05:07.775198Z 9 Query commit 2018-01-02T01:05:07.779582Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘NYCSI1HS61BN6QAVVYTZSC‘,now(),@@server_id) 2018-01-02T01:05:07.780174Z 10 Init DB replcrash 2018-01-02T01:05:07.780249Z 10 Query SAVEPOINT sp 2018-01-02T01:05:07.780913Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘NYCSI1HS61BN6QAVVYTZSC‘,now(),@@server_id) 2018-01-02T01:05:07.781387Z 9 Query commit 2018-01-02T01:05:07.781776Z 10 Query show tables 2018-01-02T01:05:07.782078Z 10 Query show table status like ‘py\_user‘ 2018-01-02T01:05:07.782400Z 10 Query SET SQL_QUOTE_SHOW_CREATE=1 2018-01-02T01:05:07.782513Z 10 Query SET SESSION character_set_results = ‘binary‘ 2018-01-02T01:05:07.787051Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘7CFC1JQPIQGNC97MDTT8ZIMIZL7D‘,now(),@@server_id) 2018-01-02T01:05:07.787810Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘7CFC1JQPIQGNC97MDTT8ZIMIZL7D‘,now(),@@server_id) 2018-01-02T01:05:07.788502Z 9 Query commit 2018-01-02T01:05:07.788774Z 10 Query show create table `py_user` 2018-01-02T01:05:07.789570Z 10 Query SET SESSION character_set_results = ‘utf8‘ 2018-01-02T01:05:07.789725Z 10 Query show fields from `py_user` 2018-01-02T01:05:07.790423Z 10 Query show fields from `py_user` 2018-01-02T01:05:07.791163Z 10 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `py_user` 2018-01-02T01:05:07.791447Z 10 Query SET SESSION character_set_results = ‘binary‘ 2018-01-02T01:05:07.791648Z 10 Query use `replcrash` 2018-01-02T01:05:07.791778Z 10 Query select @@collation_database 2018-01-02T01:05:07.791929Z 10 Query SHOW TRIGGERS LIKE ‘py\_user‘ 2018-01-02T01:05:07.792383Z 10 Query SET SESSION character_set_results = ‘utf8‘ 2018-01-02T01:05:07.792492Z 10 Query ROLLBACK TO SAVEPOINT sp 2018-01-02T01:05:07.792651Z 10 Query show table status like ‘py\_user\_innodb‘ 2018-01-02T01:05:07.792874Z 10 Query SET SQL_QUOTE_SHOW_CREATE=1 2018-01-02T01:05:07.792948Z 10 Query SET SESSION character_set_results = ‘binary‘ 2018-01-02T01:05:07.793024Z 10 Query show create table `py_user_innodb` 2018-01-02T01:05:07.793131Z 10 Query SET SESSION character_set_results = ‘utf8‘ 2018-01-02T01:05:07.793220Z 10 Query show fields from `py_user_innodb` 2018-01-02T01:05:07.793607Z 10 Query show fields from `py_user_innodb` 2018-01-02T01:05:07.793985Z 10 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `py_user_innodb` 2018-01-02T01:05:07.794435Z 9 Query insert into py_user_myisam(name,add_time,server_id) values(‘GA78AR4Z12WQTEAM41JB‘,now(),@@server_id) 2018-01-02T01:05:07.795204Z 9 Query insert into py_user_innodb(name,add_time,server_id) values(‘GA78AR4Z12WQTEAM41JB‘,now(),@@server_id) 2018-01-02T01:05:07.795688Z 9 Query commit 2018-01-02T01:05:07.798108Z 10 Query SET SESSION character_set_results = ‘binary‘ 2018-01-02T01:05:07.798205Z 10 Query use `replcrash` 2018-01-02T01:05:07.798303Z 10 Query select @@collation_database 2018-01-02T01:05:07.798408Z 10 Query SHOW TRIGGERS LIKE ‘py\_user\_innodb‘ 2018-01-02T01:05:07.798884Z 10 Query SET SESSION character_set_results = ‘utf8‘ 2018-01-02T01:05:07.798965Z 10 Query ROLLBACK TO SAVEPOINT sp 2018-01-02T01:05:07.799049Z 10 Query show table status like ‘py\_user\_myisam‘ 2018-01-02T01:05:07.799271Z 10 Query SET SQL_QUOTE_SHOW_CREATE=1 2018-01-02T01:05:07.799344Z 10 Query SET SESSION character_set_results = ‘binary‘ 2018-01-02T01:05:07.799420Z 10 Query show create table `py_user_myisam` 2018-01-02T01:05:07.799554Z 10 Query SET SESSION character_set_results = ‘utf8‘ 2018-01-02T01:05:07.799661Z 10 Query show fields from `py_user_myisam` 2018-01-02T01:05:07.800098Z 10 Query show fields from `py_user_myisam` 2018-01-02T01:05:07.800418Z 10 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `py_user_myisam`
mysqldump备份过程(--single-transaction --master-data)：
会话先执行FTWRL(实例只读)，然后设置RR隔离级别->START TRANSACTION WITH CONSISTENT SNAPSHOT;->SHOW MASTER STATUS;->UNLOCK TABLES;->SELECT /*!40001 SQL_NO_CACHE */ * FROM `tbname`;
在UNLOCK TABLES解锁后其他事务就可以进行写入操作。general-log中我们可看到 UNLOCK TABLES 到 SELECT /*!40001 SQL_NO_CACHE */ * FROM `py_user_myisam` 之间往py_user_myisam、py_user_innodb各写入9条数据
SELECT /*!40001 SQL_NO_CACHE */ * FROM `py_user_innodb`;读取的是START TRANSACTION WITH CONSISTENT SNAPSHOT建立时的数据
SELECT /*!40001 SQL_NO_CACHE */ * FROM `py_user_myisam`;读取的是最新的数据
再来查看上述过程期间binary log记录
[root@ZST1 logs]# mysqlbinlog -vv --base64-output=decode-rows mysql-bin.000183 |more ... COMMIT/*!*/; # at 1042059 #180102 9:05:07 server id 1323306 end_log_pos 1042124 CRC32 0x221cda50 GTID last_committed=2917 sequence_number=2918 rbr_only=yes /*!50718 SET TRANSACTION ISOLATION LEVEL READ COMMITTED*//*!*/; SET @@SESSION.GTID_NEXT= ‘8ab82362-9c37-11e7-a858-000c29c1025c:251873‘/*!*/; # at 1042124 #180102 9:05:07 server id 1323306 end_log_pos 1042209 CRC32 0x5df266e4 Query thread_id=9 exec_time=0 error_code=0 SET TIMESTAMP=1514855107/*!*/; BEGIN /*!*/; # at 1042209 #180102 9:05:07 server id 1323306 end_log_pos 1042279 CRC32 0xc1d41c5f Table_map: `replcrash`.`py_user_myisam` mapped to number 254 # at 1042279 #180102 9:05:07 server id 1323306 end_log_pos 1042355 CRC32 0x27badc02 Write_rows: table id 254 flags: STMT_END_F ### INSERT INTO `replcrash`.`py_user_myisam` ### SET ### @1=331 /* INT meta=0 nullable=0 is_null=0 */ ### @2=‘7ATZSNFNIBW5DZNMNZYBMV‘ /* VARSTRING(96) meta=96 nullable=1 is_null=0 */ ### @3=‘2018-01-02 09:05:07‘ /* DATETIME(0) meta=0 nullable=1 is_null=0 */ ### @4=‘1323306‘ /* VARSTRING(30) meta=30 nullable=1 is_null=0 */ # at 1042355 #180102 9:05:07 server id 1323306 end_log_pos 1042441 CRC32 0x67285443 Query thread_id=9 exec_time=0 error_code=0 SET TIMESTAMP=1514855107/*!*/; COMMIT /*!*/; # at 1042441 ==================== mysqldump备份文件获取的binlog pos ==================== ==================== 备份文件还原后，从库GTID_PURGED位置 ==================== #180102 9:05:07 server id 1323306 end_log_pos 1042506 CRC32 0xf77ede80 GTID last_committed=2918 sequence_number=2919 rbr_only=yes /*!50718 SET TRANSACTION ISOLATION LEVEL READ COMMITTED*//*!*/; SET @@SESSION.GTID_NEXT= ‘8ab82362-9c37-11e7-a858-000c29c1025c:251874‘/*!*/; # at 1042506 #180102 9:05:07 server id 1323306 end_log_pos 1042591 CRC32 0x506a2875 Query thread_id=9 exec_time=0 error_code=0 SET TIMESTAMP=1514855107/*!*/; BEGIN /*!*/; # at 1042591 #180102 9:05:07 server id 1323306 end_log_pos 1042661 CRC32 0x90b154e8 Table_map: `replcrash`.`py_user_innodb` mapped to number 255 # at 1042661 #180102 9:05:07 server id 1323306 end_log_pos 1042737 CRC32 0x1d693238 Write_rows: table id 255 flags: STMT_END_F ### INSERT INTO `replcrash`.`py_user_innodb` ### SET ### @1=331 /* INT meta=0 nullable=0 is_null=0 */ ### @2=‘7ATZSNFNIBW5DZNMNZYBMV‘ /* VARSTRING(96) meta=96 nullable=1 is_null=0 */ ### @3=‘2018-01-02 09:05:07‘ /* DATETIME(0) meta=0 nullable=1 is_null=0 */ ### @4=‘1323306‘ /* VARSTRING(30) meta=30 nullable=1 is_null=0 */ # at 1042737 #180102 9:05:07 server id 1323306 end_log_pos 1042768 CRC32 0x87864022 Xid = 4420 COMMIT/*!*/; # at 1042768 ==================== 启动复制后，py_user_innodb写入uid=331记录，成功 ==================== #180102 9:05:07 server id 1323306 end_log_pos 1042833 CRC32 0xe492578a GTID last_committed=2919 sequence_number=2920 rbr_only=yes /*!50718 SET TRANSACTION ISOLATION LEVEL READ COMMITTED*//*!*/; SET @@SESSION.GTID_NEXT= ‘8ab82362-9c37-11e7-a858-000c29c1025c:251875‘/*!*/; # at 1042833 #180102 9:05:07 server id 1323306 end_log_pos 1042918 CRC32 0xf08c4165 Query thread_id=9 exec_time=0 error_code=0 SET TIMESTAMP=1514855107/*!*/; BEGIN /*!*/; # at 1042918 #180102 9:05:07 server id 1323306 end_log_pos 1042988 CRC32 0xf16731d6 Table_map: `replcrash`.`py_user_myisam` mapped to number 257 # at 1042988 #180102 9:05:07 server id 1323306 end_log_pos 1043062 CRC32 0x128aec5e Write_rows: table id 257 flags: STMT_END_F ### INSERT INTO `replcrash`.`py_user_myisam` ### SET ### @1=332 /* INT meta=0 nullable=0 is_null=0 */ ### @2=‘X1LME9HO5V7WXNOKBVZE‘ /* VARSTRING(96) meta=96 nullable=1 is_null=0 */ ### @3=‘2018-01-02 09:05:07‘ /* DATETIME(0) meta=0 nullable=1 is_null=0 */ ### @4=‘1323306‘ /* VARSTRING(30) meta=30 nullable=1 is_null=0 */ # at 1043062 #180102 9:05:07 server id 1323306 end_log_pos 1043148 CRC32 0x5cc8cc30 Query thread_id=9 exec_time=0 error_code=0 SET TIMESTAMP=1514855107/*!*/; COMMIT /*!*/; # at 1043148 ==================== 启动复制后，py_user_myisam写入uid=332记录，失败 ==================== ==================== 从库py_user_myisam表已存在332记录，sql_thread停止 ==================== #180102 9:05:07 server id 1323306 end_log_pos 1043213 CRC32 0xceb1ce4d GTID last_committed=2920 sequence_number=2921 rbr_only=yes /*!50718 SET TRANSACTION ISOLATION LEVEL READ COMMITTED*//*!*/; SET @@SESSION.GTID_NEXT= ‘8ab82362-9c37-11e7-a858-000c29c1025c:251876‘/*!*/; # at 1043213 #180102 9:05:07 server id 1323306 end_log_pos 1043298 CRC32 0x38591b71 Query thread_id=9 exec_time=0 error_code=0 SET TIMESTAMP=1514855107/*!*/; BEGIN /*!*/; # at 1043298 #180102 9:05:07 server id 1323306 end_log_pos 1043368 CRC32 0xf80c2ae9 Table_map: `replcrash`.`py_user_innodb` mapped to number 258 # at 1043368 #180102 9:05:07 server id 1323306 end_log_pos 1043442 CRC32 0x0bf4ae26 Write_rows: table id 258 flags: STMT_END_F ### INSERT INTO `replcrash`.`py_user_innodb` ### SET ### @1=332 /* INT meta=0 nullable=0 is_null=0 */ ### @2=‘X1LME9HO5V7WXNOKBVZE‘ /* VARSTRING(96) meta=96 nullable=1 is_null=0 */ ### @3=‘2018-01-02 09:05:07‘ /* DATETIME(0) meta=0 nullable=1 is_null=0 */ ### @4=‘1323306‘ /* VARSTRING(30) meta=30 nullable=1 is_null=0 */ # at 1043442 #180102 9:05:07 server id 1323306 end_log_pos 1043473 CRC32 0x1b75f9e0 Xid = 4431 COMMIT/*!*/; # at 1043473 ...
二进志日志显示，按照py_user_myisam->py_user_innodb->py_user_myisam这样的顺序往表中写入数据。
使用备份文件搭建的从库，py_user_myisam表已包含UNLOCK TABLES之后的9条数据，但在备份文件中的@@GLOBAL.GTID_PURGED却是UNLOCK TABLES时刻的位置。因此在启动复制后，第一条操作py_user_innodb表成功，第二条操作py_user_myisam表失败，从库报主键冲突错误
那么应该如何修复这个错误呢?
# 删除从库py_user_myisam表大于等于冲突key的记录 mydba@192.168.85.132,3308 [replcrash]> delete from py_user_myisam where uid>=332; # 重新启动sql_thread mydba@192.168.85.132,3308 [replcrash]> start slave sql_thread;
总得来说就是只有innodb才会提供一致性备份！！！
1.2、DDL操作对备份的影响
建议先阅读后续章节，理解各备份过程后再返回阅读DDL操作对备份的影响
运行下面的脚本持续DDL操作
#!/user/bin/python import string import random import MySQLdb import time conn = MySQLdb.connect(host=‘192.168.85.132‘, port=3306, user=‘mydba‘, passwd=‘mysql5719‘, db=‘replcrash‘) """ create table py_user( uid int not null auto_increment, name varchar(32), add_time datetime default current_timestamp, server_id varchar(10), primary key(uid), key(name) ); """ counter = 1 while counter<=100: addcol = ‘col‘ + str(counter) print addcol cursor = conn.cursor() #DDL cursor.execute("alter table py_user_innodb add %s int;" % addcol) cursor.execute("alter table py_user_innodb drop column %s;" % addcol) conn.commit() counter += 1 time.sleep(0.0001)
逻辑备份数据库
# 清空general_log [root@ZST1 logs]# cat /dev/null > /data/mysql/mysql3306/data/mysql-general.log # mysqldump [root@ZST1 backup]# mysqldump -h127.0.0.1 -P3306 -uroot -p --single-transaction --master-data=2 replcrash >/data/backup/replcrash_dump_1323306_`date +%Y%m%d`.sql Enter password: Warning: A partial dump from a server that has GTIDs will by default include the GTIDs of all transactions, even those that changed suppressed parts of the database. If you don not want to restore GTIDs, pass --set-gtid-purged=OFF. To make a complete dump, pass --all-databases --triggers --routines --events. mysqldump: Error 1412: Table definition has changed, please retry transaction when dumping table `py_user_innodb` at row: 0 # mysqlpump [root@ZST1 backup]# mysqlpump -h127.0.0.1 -P3306 -uroot -p --single-transaction --add-drop-table --exclude-databases=mysql,sakila,backupdb -A >/data/backup/replcrash_pump_1323306_`date +%Y%m%d`.sql Enter password: mysqlpump: [ERROR] (1412) Table definition has changed, please retry transaction Dump process encountered error and will not continue. Dump progress: 0/3 tables, 250/2431 rows [root@ZST1 backup]# # mydumper [root@ZST1 mydumper]# mydumper -h 127.0.0.1 -P 3306 -u root -p mysql5719 --trx-consistency-only -v 3 -t 2 -o /data/backup/mydumper .. ** Message: Thread 2 dumping data for `replcrash`.`py_user_innodb` ** Message: Thread 1 dumping data for `replcrash`.`py_user_myisam` ** (mydumper:5179): CRITICAL **: Could not read data from replcrash.py_user_innodb: Table definition has changed, please retry transaction ** Message: Empty table replcrash.py_user_innodb ...
mydumper需要使用-v 3 显示详细信息，本身是不会报错的！！！如果想更容易再现错误，最好是备份的数据表较大、较多，适当降低并行线程数(-t)，开启--trx-consistency-only，让其尽早解锁
生成的备份文件中，py_user_innodb表只有结构，没有数据
根据备份逻辑，在UNLOCK TABLES解除FTWRL到SELECT /*!40001 SQL_NO_CACHE */ * FROM `tbname`之间如果有DDL操作，就会造成上述错误
使用了with consistent snapshot子句开启一致性快照事务之后，如果一旦表结构定义发生改变，事务将无法对该表执行查询
| 使用WITH CONSISTENT SNAPSHOT子句，会话1显式开启一个事务之后先不执行查询，会话B使用DDL语句添加一个字段 | |
| 会话1 | 会话2 |
|
修改隔离级别为RR |
修改隔离级别为RR |
|
显式开启一个事务，先不执行查询 |
|
|
执行DDL语句添加字段，执行成功 |
|
|
执行查询，报表定义已经改变的错误 |
|
物理备份数据库
[root@ZST1 backup]# innobackupex --defaults-file=/data/mysql/mysql3306/my.cnf -S /tmp/mysql3306.sock -uroot -pmysql5719 /data/backup/full/ 180104 15:29:31 [01] Copying ./sakila/payment.ibd to /data/backup/full/2018-01-04_15-29-21/sakila/payment.ibd 180104 15:29:32 [01] ...done InnoDB: Last flushed lsn: 19337348200 load_index lsn 19337354022 InnoDB: An optimized (without redo logging) DDLoperation has been performed. All modified pages may not have been flushed to the disk yet. PXB will not be able take a consistent backup. Retry the backup operation 180104 15:29:32 [01] Copying ./replcrash/py_user.ibd to /data/backup/full/2018-01-04_15-29-21/replcrash/py_user.ibd 180104 15:29:32 [01] ...done 180104 15:29:32 [01] Copying ./replcrash/py_user_innodb.ibd to /data/backup/full/2018-01-04_15-29-21/replcrash/py_user_innodb.ibd 180104 15:29:32 [01] ...done
只要在备份期间(实际是备份InnoDB表期间，因为备份non-InnoDB表期间会加FTWRL只读锁，阻塞DDL、DML操作)执行DDL操作，innobackupex就会报错退出。原因就是DDL操作不会记录到redo log，PXB will not be able take a consistent backup.
因此备份期间要避免执行不记录事务日志的操作(ALTER TABLE, CREATE TABLE, DROP TABLE, RENAME TABLE, TRUNCATE TABLE)
二、mysqlpump
2.1、备份过程
借助前面已开启general_log，来查看mysqlpump执行过程
MySQL 5.7.11起解决了--single-transaction和--default-parallelism的互斥问题
# 清空general_log [root@ZST1 logs]# cat /dev/null > /data/mysql/mysql3306/data/mysql-general.log # 备份replcrash数据库，一致性备份(源码限制只有备份整个实例时才能返回GTID信息，因此这里使用-A，然后使用--exclude-databases排除不需要备份的db) [root@ZST1 backup]# mysqlpump -h127.0.0.1 -P3306 -uroot -p --single-transaction --add-drop-table --exclude-databases=mysql,sakila,backupdb -A >/data/backup/replcrash_pump_1323306_`date +%Y%m%d`.sql 默认mysqlpump使用一个队列两个线程 [root@ZST1 data]# vim /data/mysql/mysql3306/data/mysql-general.log ... 4 Time Id Command Argument 5 2018-01-03T01:24:06.623704Z 14 Connect root@localhost on using TCP/IP 6 2018-01-03T01:24:06.631266Z 14 Query FLUSH TABLES WITH READ LOCK 7 2018-01-03T01:24:06.667093Z 14 Query SHOW WARNINGS 8 2018-01-03T01:24:06.667310Z 14 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 9 2018-01-03T01:24:06.667520Z 14 Query SHOW WARNINGS 10 2018-01-03T01:24:06.667647Z 14 Query START TRANSACTION WITH CONSISTENT SNAPSHOT 11 2018-01-03T01:24:06.667792Z 14 Query SHOW WARNINGS 12 2018-01-03T01:24:06.679491Z 15 Connect root@localhost on using TCP/IP 13 2018-01-03T01:24:06.683019Z 15 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 14 2018-01-03T01:24:06.684865Z 15 Query SHOW WARNINGS 15 2018-01-03T01:24:06.685015Z 15 Query START TRANSACTION WITH CONSISTENT SNAPSHOT 16 2018-01-03T01:24:06.685114Z 15 Query SHOW WARNINGS 17 2018-01-03T01:24:06.686057Z 16 Connect root@localhost on using TCP/IP 18 2018-01-03T01:24:06.688856Z 16 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 19 2018-01-03T01:24:06.716231Z 16 Query SHOW WARNINGS 20 2018-01-03T01:24:06.716447Z 16 Query START TRANSACTION WITH CONSISTENT SNAPSHOT 21 2018-01-03T01:24:06.716558Z 16 Query SHOW WARNINGS 22 2018-01-03T01:24:06.716701Z 14 Query UNLOCK TABLES 23 2018-01-03T01:24:06.716857Z 14 Query SHOW WARNINGS ...各种SHOW... 1176 2018-01-03T01:24:07.344380Z 16 Query SHOW CREATE DATABASE IF NOT EXISTS `replcrash` 1177 2018-01-03T01:24:07.344468Z 16 Query SHOW WARNINGS 1178 2018-01-03T01:24:07.344565Z 16 Query SHOW TABLE STATUS FROM `replcrash` 1179 2018-01-03T01:24:07.380209Z 16 Query SHOW WARNINGS 1180 2018-01-03T01:24:07.380416Z 16 Query SHOW COLUMNS IN `py_user` FROM `replcrash` 1181 2018-01-03T01:24:07.381223Z 16 Query SHOW WARNINGS 1182 2018-01-03T01:24:07.381408Z 16 Query SHOW CREATE TABLE `replcrash`.`py_user` 1183 2018-01-03T01:24:07.381614Z 16 Query SHOW WARNINGS 1184 2018-01-03T01:24:07.381950Z 16 Query SHOW TRIGGERS FROM `replcrash` LIKE ‘py_user‘ 1185 2018-01-03T01:24:07.382575Z 16 Query SHOW WARNINGS 1186 2018-01-03T01:24:07.382764Z 16 Query SHOW COLUMNS IN `py_user_innodb` FROM `replcrash` 1187 2018-01-03T01:24:07.383125Z 14 Query SET SQL_QUOTE_SHOW_CREATE= 1 1188 2018-01-03T01:24:07.383334Z 14 Query SHOW WARNINGS 1189 2018-01-03T01:24:07.383617Z 14 Query SET TIME_ZONE=‘+00:00‘ 1190 2018-01-03T01:24:07.384037Z 14 Query SHOW WARNINGS 1191 2018-01-03T01:24:07.385106Z 15 Query SELECT `COLUMN_NAME`, `EXTRA` FROM `INFORMATION_SCHEMA`.`COLUMNS`WHERE TABLE_SCHEMA =‘replcrash‘ AND TABLE_NAME =‘py_user‘ 1192 2018-01-03T01:24:07.386099Z 15 Query SHOW WARNINGS 1193 2018-01-03T01:24:07.386347Z 15 Query SELECT SQL_NO_CACHE `uid`,`name`,`add_time`,`server_id` FROM `replcrash`.`py_user` 1194 2018-01-03T01:24:07.387102Z 15 Query SHOW WARNINGS 1195 2018-01-03T01:24:07.387644Z 16 Query SHOW WARNINGS 1196 2018-01-03T01:24:07.387997Z 16 Query SHOW CREATE TABLE `replcrash`.`py_user_innodb` 1197 2018-01-03T01:24:07.388216Z 16 Query SHOW WARNINGS 1198 2018-01-03T01:24:07.388487Z 16 Query SHOW TRIGGERS FROM `replcrash` LIKE ‘py_user_innodb‘ 1199 2018-01-03T01:24:07.389053Z 14 Query SELECT `COLUMN_NAME`, `EXTRA` FROM `INFORMATION_SCHEMA`.`COLUMNS`WHERE TABLE_SCHEMA =‘replcrash‘ AND TABLE_NAME =‘py_user_innodb‘ 1200 2018-01-03T01:24:07.390054Z 14 Query SHOW WARNINGS 1201 2018-01-03T01:24:07.390293Z 14 Query SELECT SQL_NO_CACHE `uid`,`name`,`add_time`,`server_id` FROM `replcrash`.`py_user_innodb` 1202 2018-01-03T01:24:07.391566Z 16 Query SHOW WARNINGS 1203 2018-01-03T01:24:07.391776Z 16 Query SHOW COLUMNS IN `py_user_myisam` FROM `replcrash` 1204 2018-01-03T01:24:07.392559Z 16 Query SHOW WARNINGS 1205 2018-01-03T01:24:07.392747Z 16 Query SHOW CREATE TABLE `replcrash`.`py_user_myisam` 1206 2018-01-03T01:24:07.393065Z 16 Query SHOW WARNINGS 1207 2018-01-03T01:24:07.393336Z 16 Query SHOW TRIGGERS FROM `replcrash` LIKE ‘py_user_myisam‘ 1208 2018-01-03T01:24:07.394146Z 16 Query SHOW WARNINGS 1209 2018-01-03T01:24:07.394371Z 16 Query SHOW FUNCTION STATUS WHERE db = ‘replcrash‘ 1210 2018-01-03T01:24:07.396083Z 15 Query SELECT `COLUMN_NAME`, `EXTRA` FROM `INFORMATION_SCHEMA`.`COLUMNS`WHERE TABLE_SCHEMA =‘replcrash‘ AND TABLE_NAME =‘py_user_myisam‘ 1211 2018-01-03T01:24:07.399053Z 15 Query SHOW WARNINGS 1212 2018-01-03T01:24:07.399425Z 15 Query SELECT SQL_NO_CACHE `uid`,`name`,`add_time`,`server_id` FROM `replcrash`.`py_user_myisam` 1213 2018-01-03T01:24:07.405719Z 16 Query SHOW WARNINGS 1214 2018-01-03T01:24:07.405915Z 16 Query SHOW PROCEDURE STATUS WHERE db = ‘replcrash‘ 1215 2018-01-03T01:24:07.412340Z 14 Query SHOW WARNINGS 1216 2018-01-03T01:24:07.414496Z 15 Query SHOW WARNINGS 1217 2018-01-03T01:24:07.416755Z 16 Query SHOW WARNINGS 1218 2018-01-03T01:24:07.417261Z 16 Query SHOW EVENTS FROM `replcrash` 1219 2018-01-03T01:24:07.417884Z 16 Query SHOW WARNINGS
即使备份一个db，general_log中也会出现很多其他库的内容(⊙_⊙)
mysqlpump备份过程(--single-transaction)：
对于建立的第一个连接，执行FLUSH TABLES WITH READ LOCK，加上只读锁；对于其他连接(包含第一个连接)，设置RR隔离级别，并开启一致性快照读START TRANSACTION WITH CONSISTENT SNAPSHOT；当所有连接(--default-parallelism设置多少个线程)都建立好后，再执行解锁UNLOCK TABLES；最后通过SELECT colname FROM `tbname`备份数据
general_log中没有看到SHOW MASTER STATUS，猜测它类似于mydumper，也是在主线程FLUSH TABLES WITH READ LOCK后，就去获取GTID信息。
因此mysqlpump也会遇到在mysqldump中的坑
mysqlpump并行备份的部分参数需要在单线程模式才能应用，实际使用中建议先测试，对比输出结果分辨各参数的作用
2.2、--single-transaction
• --single-transaction
This option sets the transaction isolation mode to REPEATABLE READ and sends a START TRANSACTION SQL statement to the server before dumping data. It is useful only with transactional tables such as InnoDB, because then it dumps the consistent state of the database at the time when START TRANSACTION was issued without blocking any applications.
When using this option, you should keep in mind that only InnoDB tables are dumped in a consistent state. For example, any MyISAM or MEMORY tables dumped while using this option may still change state.
While a --single-transaction dump is in process, to ensure a valid dump file (correct table contents and binary log coordinates), no other connection should use the following statements: ALTER TABLE, CREATE TABLE, DROP TABLE, RENAME TABLE, TRUNCATE TABLE. A consistent read is not isolated from those statements, so use of them on a table to be dumped can cause the SELECT that is performed by mysqldump/mysqlpump to retrieve the table contents to obtain incorrect contents or fail.
• START TRANSACTION
The WITH CONSISTENT SNAPSHOT modifier starts a consistent read for storage engines that are capable of it. This applies only to InnoDB. The effect is the same as issuing a START TRANSACTION followed by a SELECT from any InnoDB table.
The WITH CONSISTENT SNAPSHOT modifier does not change the current transaction isolation level, so it provides a consistent snapshot only if the current isolation level is one that permits a consistent read. The only isolation level that permits a consistent read is REPEATABLE READ.
该参数将事务隔离级别设置成Repeatable Read，并在dump之前发送start transaction语句给服务端。这只对事务表(比如innodb)很有用，因为在发出start transaction时，保证了在不阻塞任何应用下的一致性状态。对myisam和memory等非事务表，还是会改变状态的，当使用此参的时候要确保没有其他连接在使用ALTER TABLE、CREATE TABLE、DROP TABLE、RENAME TABLE、TRUNCATE TABLE等语句，否则会出现不正确的内容或者失败。在mysql5.7.11之前，--default-parallelism大于1的时候和此参数互斥，必须使用--default-parallelism=0。5.7.11之后解决了--single-transaction和--default-parallelism的互斥问题
三、mydumper
3.1、备份过程
mydumper备份过程
1、连接目标数据库
2、通过show processlist来判断是否有长查询，根据参数long-query-guard和kill-long-queries决定退出或杀掉长查询
3、主线程flush tables with read lock；start transaction with consistent snapshot
4、主线程读取当前时间点的二进制日志文件名和日志写入的位置并记录在metadata文件中，以供即时点恢复使用
5、创建dump线程，缺省为4个
6、确定候选表，根据类别分别插入non_innodb_table、innodb_tables以及table_schemas链表
7、将候选表通过g_async_queue_push加入任务队列(队列最后元素是thread shutdown)，由dump线程从队列中读取表信息并执行数据导出
8、备份完non_innodb_table后立即unlock tables解锁，以减少锁定时间
9、等待dump InnoDB tables完成
从备份逻辑中可以看出，mydumper需要在备份完非事务表之后才解锁，这就保证了事务表和非事务表的一致性备份。默认情况下，DML操作不会影响mydumper备份的一致性
3.2、--trx-consistency-only
这个参数退化为仅保证事务表的一致性备份，在备份非事务表前如果有数据写入就会导致数据与metadata不一致，有点类似前面的mysqldump、mysqlpump的味道
运行前面的PythonDML脚本，一直往py_user_myisam、py_user_innodb表中写入数据，测试--trx-consistency-only选项的效果
# 清空general_log [root@ZST1 logs]# cat /dev/null > /data/mysql/mysql3306/data/mysql-general.log # 备份replcrash库(仅事务表一致性备份) [root@ZST1 mydumper]# mydumper -h 127.0.0.1 -P 3306 -u root -p mysql5719 --trx-consistency-only -v 3 -B replcrash -o /data/backup/mydumper ** (mydumper:2808): WARNING **: Using trx_consistency_only, binlog coordinates will not be accurate if you are writing to non transactional tables. ** Message: Connected to a MySQL server ** Message: Started dump at: 2018-01-04 09:34:53 ** Message: Written master status ** Message: Thread 1 connected using MySQL connection ID 50 ** Message: Thread 2 connected using MySQL connection ID 51 ** Message: Thread 3 connected using MySQL connection ID 52 ** Message: Thread 4 connected using MySQL connection ID 53 ** Message: Transactions started, unlocking tables ** Message: Thread 1 dumping data for `replcrash`.`py_user` ** Message: Thread 2 dumping data for `replcrash`.`py_user_innodb` ** Message: Thread 2 dumping data for `replcrash`.`py_user_myisam` ** Message: Thread 1 dumping schema for `replcrash`.`py_user` ** Message: Thread 3 dumping schema for `replcrash`.`py_user_innodb` ** Message: Thread 1 dumping schema for `replcrash`.`py_user_myisam` ** Message: Thread 3 shutting down ** Message: Thread 1 shutting down ** Message: Thread 2 shutting down ** Message: Thread 4 shutting down ** Message: Finished dump at: 2018-01-04 09:34:53 # 查看metadata [root@ZST1 mydumper]# cat metadata Started dump at: 2018-01-04 09:34:53 SHOW MASTER STATUS: Log: mysql-bin.000185 Pos: 2452139 GTID:8ab82362-9c37-11e7-a858-000c29c1025c:1-272303 Finished dump at: 2018-01-04 09:34:53 [root@ZST1 mydumper]# # 还原replcrash库 [root@ZST1 mydumper]# myloader -h 127.0.0.1 -P 3308 -u root -p mysql5719 -o -B replcrash -d /data/backup/mydumper myloader不会产生binlog，也不会应用metadata中的GTID # 设置GTID_PURGED mydba@192.168.85.132,3308 [replcrash]> reset master; mydba@192.168.85.132,3308 [replcrash]> SET @@GLOBAL.GTID_PURGED=‘8ab82362-9c37-11e7-a858-000c29c1025c:1-272303‘; # 搭建复制 mydba@192.168.85.132,3308 [replcrash]> change master to master_host=‘192.168.85.132‘, master_port=3306, master_user=‘repl‘, master_password=‘repl‘, master_auto_position=1; # 启动复制，查看复制状态 mydba@192.168.85.132,3308 [replcrash]> start slave; mydba@192.168.85.132,3308 [replcrash]> show slave status\G *************************** 1. row *************************** Master_Log_File: mysql-bin.000185 Read_Master_Log_Pos: 2758886 Relay_Log_File: relay-bin.000002 Relay_Log_Pos: 414 Relay_Master_Log_File: mysql-bin.000185 Slave_IO_Running: Yes Slave_SQL_Running: No Exec_Master_Log_Pos: 2452139 Last_SQL_Errno: 1062 Last_SQL_Error: Could not execute Write_rows event on table replcrash.py_user_myisam; Duplicate entry ‘783‘ for key ‘PRIMARY‘, Error_code: 1062; handler error HA_ERR_FOUND_DUPP_KEY; the event‘‘s master log mysql-bin.000185, end_log_pos 2452443 Retrieved_Gtid_Set: 8ab82362-9c37-11e7-a858-000c29c1025c:272304-273161 Executed_Gtid_Set: 8ab82362-9c37-11e7-a858-000c29c1025c:1-272303 Auto_Position: 1 # 大于等于冲突key的数据 mydba@192.168.85.132,3308 [replcrash]> select * from replcrash.py_user_myisam where uid>=783; +-----+--------------------------------+---------------------+-----------+ | uid | name | add_time | server_id | +-----+--------------------------------+---------------------+-----------+ | 783 | MLGU22VB26RHNNYAY6IPPUJX9A74EM | 2018-01-04 09:34:53 | 1323306 | +-----+--------------------------------+---------------------+-----------+ 1 row in set (0.06 sec) # mydumper产生的general-log [root@ZST1 data]# vim /data/mysql/mysql3306/data/mysql-general.log ... 2018-01-04T01:34:53.769877Z 49 Query SET SESSION net_write_timeout = 2147483 2018-01-04T01:34:53.770302Z 49 Query SHOW PROCESSLIST 2018-01-04T01:34:53.770886Z 49 Query FLUSH TABLES WITH READ LOCK 2018-01-04T01:34:53.771236Z 49 Query START TRANSACTION /*!40108 WITH CONSISTENT SNAPSHOT */ 2018-01-04T01:34:53.771430Z 49 Query /*!40101 SET NAMES binary*/ 2018-01-04T01:34:53.772084Z 49 Query SHOW MASTER STATUS 2018-01-04T01:34:53.772331Z 49 Query SHOW SLAVE STATUS 2018-01-04T01:34:53.773386Z 48 Query insert into py_user_myisam(name,add_time,server_id) values(‘MLGU22VB26RHNNYAY6IPPUJX9A74EM‘,now(),@@server_id) 2018-01-04T01:34:53.773981Z 50 Connect root@localhost on using TCP/IP 2018-01-04T01:34:53.788273Z 50 Query SET SESSION wait_timeout = 2147483 2018-01-04T01:34:53.788874Z 50 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 2018-01-04T01:34:53.788972Z 50 Query START TRANSACTION /*!40108 WITH CONSISTENT SNAPSHOT */ 2018-01-04T01:34:53.789092Z 50 Query /*!40103 SET TIME_ZONE=‘+00:00‘ */ 2018-01-04T01:34:53.789201Z 50 Query /*!40101 SET NAMES binary*/ 2018-01-04T01:34:53.790091Z 51 Connect root@localhost on using TCP/IP 2018-01-04T01:34:53.790315Z 51 Query SET SESSION wait_timeout = 2147483 2018-01-04T01:34:53.791912Z 51 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 2018-01-04T01:34:53.792095Z 51 Query START TRANSACTION /*!40108 WITH CONSISTENT SNAPSHOT */ 2018-01-04T01:34:53.792307Z 51 Query /*!40103 SET TIME_ZONE=‘+00:00‘ */ 2018-01-04T01:34:53.792418Z 51 Query /*!40101 SET NAMES binary*/ 2018-01-04T01:34:53.793381Z 52 Connect root@localhost on using TCP/IP 2018-01-04T01:34:53.794631Z 52 Query SET SESSION wait_timeout = 2147483 2018-01-04T01:34:53.796301Z 52 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 2018-01-04T01:34:53.796404Z 52 Query START TRANSACTION /*!40108 WITH CONSISTENT SNAPSHOT */ 2018-01-04T01:34:53.796538Z 52 Query /*!40103 SET TIME_ZONE=‘+00:00‘ */ 2018-01-04T01:34:53.796641Z 52 Query /*!40101 SET NAMES binary*/ 2018-01-04T01:34:53.797414Z 53 Connect root@localhost on using TCP/IP 2018-01-04T01:34:53.798420Z 53 Query SET SESSION wait_timeout = 2147483 2018-01-04T01:34:53.799316Z 53 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ 2018-01-04T01:34:53.799425Z 53 Query START TRANSACTION /*!40108 WITH CONSISTENT SNAPSHOT */ 2018-01-04T01:34:53.800345Z 53 Query /*!40103 SET TIME_ZONE=‘+00:00‘ */ 2018-01-04T01:34:53.800449Z 53 Query /*!40101 SET NAMES binary*/ 2018-01-04T01:34:53.800661Z 49 Query UNLOCK TABLES /* trx-only */ 2018-01-04T01:34:53.801868Z 49 Init DB replcrash 2018-01-04T01:34:53.802332Z 49 Query SHOW TABLE STATUS 2018-01-04T01:34:53.804759Z 50 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `replcrash`.`py_user` 2018-01-04T01:34:53.805137Z 51 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `replcrash`.`py_user_innodb` 2018-01-04T01:34:53.807164Z 51 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `replcrash`.`py_user_myisam` 2018-01-04T01:34:53.808786Z 48 Query insert into py_user_innodb(name,add_time,server_id) values(‘MLGU22VB26RHNNYAY6IPPUJX9A74EM‘,now(),@@server_id) 2018-01-04T01:34:53.809449Z 49 Query SHOW CREATE DATABASE `replcrash` 2018-01-04T01:34:53.810326Z 52 Query SHOW CREATE TABLE `replcrash`.`py_user_innodb` 2018-01-04T01:34:53.810638Z 50 Query SHOW CREATE TABLE `replcrash`.`py_user` 2018-01-04T01:34:53.811307Z 52 Quit 2018-01-04T01:34:53.811506Z 50 Query SHOW CREATE TABLE `replcrash`.`py_user_myisam` 2018-01-04T01:34:53.812272Z 50 Quit 2018-01-04T01:34:53.812512Z 48 Query commit
48是python持续写入的线程、49是主线程、50~53是dump线程
所有连接建立后，主线程立即解锁UNLOCK TABLES /* trx-only */。主线程SHOW MASTER STATUS的后面紧接着python写入nsert into py_user_myisam，实际这个写入操作应该是在主线程UNLOCK TABLES之后，dump线程备份py_user_myisam数据之前。这就导致备份的数据与metadata不一致，因此不推荐使用--trx-consistency-only选项
3.3、逻辑备份差异
mysqldump：只支持单线程工作，这就使得它无法迅速的备份数据
mysqlpump：并行的最小粒度是单个数据库对象，对于每张表的导出只能是单个线程的
mydumper：支持对单表多个线程备份，参数-r
多线程操作提升空间受限于磁盘的IO能力，在使用前做好磁盘IO的评估
四、XtraBackup
4.1、备份过程
innobackupex全备过程
1、start xtrabackup_log
2、copy .ibd、ibdata1
3、FLUSH TABLES WITH READ LOCK
4、copy .frm、.MYD、.MYI、misc files
5、SHOW MASTER STATUS
6、UNLOCK TABLES
7、stop and copy xtrabackup_log
备份开始时首先会开启一个后台检测进程，从当前checkpoint位置开始拷贝redo log，同时持续检测redo log，一旦发现redo中有新的日志写入，立刻将日志记入后台日志文件xtrabackup_log中。之后拷贝innodb的数据文件和系统表空间文件ibdata1，待拷贝结束后，执行flush tables with read lock操作，拷贝.frm，MYI，MYD，等文件，并且在这一时刻获得binlog的位置，最后会发出unlock tables，把表设置为可读可写状态，最终停止xtrabackup_log。
InnoDB表copy [.ibd、ibdata]+[redo log]借助InnoDB Crash Recovery机制保证一致性；non-InnoDB表使用FTWRL加只读锁后backup non-InnoDB tables and files。DML操作不会影响innobackupex备份的一致性
4.2、总结
| mysqldump | mysqlpump | mydumper | innobackupex | |
| FTWRL被阻塞 | 存在互斥锁，会被阻塞 | 存在互斥锁，会被阻塞 | 存在互斥锁，会被阻塞 | 存在互斥锁，会被阻塞 |
| DDL操作导致备份异常 | 报错，表定义变更 | 报错，表定义变更 | 不报错，-v 3查看 | 报错，DDL操作没写redo log |
| DML操作影响数据一致性 | non_InnoDB表不一致 | non_InnoDB表不一致 | 默认一致，--trx-consistency-only会导致不一致 | 一致 |
| 其他 | 没有指定my.cnf导致备份一个错误的实例 |
对生产库的DDL操作、大事务、或者长时间锁表的操作，一定要避开备份时间(?ω?)
五、参考文档
mysqldump与innobackupex备份过程你知多少：http://www.cnblogs.com/xiaoboluo768/p/7560105.html
mysqlpump两个疑问点：http://www.fordba.com/mysqlpump_some_questions.html
MySQL 5.7 mysqlpump 备份工具说明：http://www.cnblogs.com/zhoujinyi/p/5684903.html
mysqlpump — A Database Backup Program：https://dev.mysql.com/doc/refman/5.7/en/mysqlpump.html
mydumper：https://github.com/maxbube/mydumper
MySQL备份之【mydumper 学习】：http://www.cnblogs.com/zhoujinyi/p/3423641.html
mydumper备份数据库详解(已详细说明)：https://yq.aliyun.com/articles/45741
mydumper备份原理和使用方法：http://www.cnblogs.com/linuxnote/p/3817698.html
