mysqldump
运行mysqldump需一定的权限。如，备份表的最低权限为select，备份触发器需show triggers权限。
（1）备份结果文件命令规范：dbname_port_$(date +%Y%m%d).bak
（2）gzip,tar是单线程压缩软件，只能用到一个cpu，效率比较低，备份的时候不建议压缩，如果空间实在不足，可以考录备份完成之后使用pigz多线程压缩软件；
（3）mysqldump是单线程工作，效率比较低，如果mysqldump备份需要很长时间，可以考虑使用mydumper支持多线程并发导出；
备份的参数
mysql>create table t1(c1 int,c2 varchar(10)); mysql>insert into t1 values(1,'aaa'),(2,'bbb'),(3,'ccc'); 例1： #指定备份单个库testdb: [root@Darren2 tmp]# mysqldump -uroot -p147258 testdb > /tmp/testdb1.bak [root@Darren2 tmp]# vim testdb1.bak SET @MYSQLDUMP_TEMP_LOG_BIN = @@SESSION.SQL_LOG_BIN; SET @@SESSION.SQL_LOG_BIN= 0; --当还原的时候不记录binlog日志 SET @@GLOBAL.GTID_PURGED='83373570-fe03-11e6-bb0a-000c29c1b8a9:1-10914'; DROP TABLE IF EXISTS `t1`; CREATE TABLE `t1` ( `c1` int(11) DEFAULT NULL, `c2` varchar(10) DEFAULT NULL ) ENGINE=InnoDB DEFAULT CHARSET=utf8; LOCK TABLES `t1` WRITE; --还原表的时候不允许其他会话读写t1表 INSERT INTO `t1` VALUES (1,'aaa'),(2,'bbb'),(3,'ccc'); UNLOCK TABLES; SET @@SESSION.SQL_LOG_BIN = @MYSQLDUMP_TEMP_LOG_BIN; 在本数据库上还原： [root@Darren2 tmp]# mysql -uroot -p147258 < testdb1.bak ERROR 1840 (HY000) at line 24: @@GLOBAL.GTID_PURGED can only be set when @@GLOBAL.GTID_EXECUTED is empty. 实质报错： root@localhost [testdb]>SET @@GLOBAL.GTID_PURGED='83373570-fe03-11e6-bb0a-000c29c1b8a9:1-10914'; ERROR 1840 (HY000): @@GLOBAL.GTID_PURGED can only be set when @@GLOBAL.GTID_EXECUTED is empty. 如果直接还原会报错，因为我开启了gtid_mode,此时可以show master status看一下executed_gtid_set参数不为空，需要在备份文件testdb1.bak中把 “SET @@GLOBAL.GTID_PURGED='83373570-fe03-11e6-bb0a-000c29c1b8a9:1-10914';”注释掉能继续还原 如果希望能够传写入binlog,把还原也同步到从库，需要注释掉SET @@SESSION.SQL_LOG_BIN= 0; root@localhost [testdb]>show master status; +------------------+----------+--------------+------------------+----------------------------------------------+ | File | Position | Binlog_Do_DB | Binlog_Ignore_DB | Executed_Gtid_Set | +------------------+----------+--------------+------------------+----------------------------------------------+ | mysql-bin.000004 | 6392 | | | 83373570-fe03-11e6-bb0a-000c29c1b8a9:1-10931 | +------------------+----------+--------------+------------------+----------------------------------------------+ 还原命令： [root@Darren2 tmp]# mysql -uroot -p147258 testdb < testdb1.bak #必须指定testdb库,否则报错找不到，如果还原testdb库之前被删除了，也会报错： [root@Darren2 tmp]# mysql -uroot -p147258 testdb <testdb1.bak mysql: [Warning] Using a password on the command line interface can be insecure. ERROR 1049 (42000): Unknown database 'testdb' 例2： 同时备份多个库-B,并且增加了create database testdb和use testdb的功能 [root@Darren2 tmp]# mysqldump -uroot -p147258 -B testdb > testdb2.bak [root@Darren2 tmp]# vim testdb2.bak SET @MYSQLDUMP_TEMP_LOG_BIN = @@SESSION.SQL_LOG_BIN; SET @@SESSION.SQL_LOG_BIN= 0; SET @@GLOBAL.GTID_PURGED='83373570-fe03-11e6-bb0a-000c29c1b8a9:1-10931'; CREATE DATABASE `testdb` ; USE `testdb`; DROP TABLE IF EXISTS `t1`; CREATE TABLE `t1` ( `c1` int(11) DEFAULT NULL, `c2` varchar(10) DEFAULT NULL ) ENGINE=InnoDB DEFAULT CHARSET=utf8; LOCK TABLES `t1` WRITE; INSERT INTO `t1` VALUES (1,'aaa'),(2,'bbb'),(3,'ccc'); UNLOCK TABLES; SET @@SESSION.SQL_LOG_BIN = @MYSQLDUMP_TEMP_LOG_BIN; 还原：无需指定testdb库 [root@Darren2 tmp]# mysql -uroot -p147258 < testdb2.bak #同时备份多个库： [root@Darren2 tmp]# mysqldump -uroot -p147258 -B testdb mysql > testdbmysql.bak 还原多个库： [root@Darren2 tmp]# mysql -uroot -p147258 < testdbmysql.bak 例3： 压缩备份 [root@Darren2 logs]# mysqldump -uroot -p147258 testdb | gzip > testdbgzip.bak.gz 还原： [root@Darren2 tmp]# gunzip < testdbgzip.bak.gz | mysql -uroot -p147258 例4： 只备份库中的表 格式：mysqldump -uroot -p147258 库名 表名1 表名2 ... >备份文件名 注意：不能加-B 参数，否则就是备份多个库了 #备份单个表： [root@Darren2 tmp]# mysqldump -uroot -p147258 testdb t1 > testdbt1.bak #备份多个表： [root@Darren2 tmp]# mysqldump -uroot -p147258 testdb t1 t2 > testdbt1t2.bak 例5： --master-data，指定日志文件从哪个位置开始，不用切割binlog日志了 [root@www ~]# mysqldump -uroot -p147258 --master-data=1 testdb CHANGE MASTER TO MASTER_LOG_FILE='mysql-bin.000011', MASTER_LOG_POS=107; [root@www opt]# mysqldump -uroot -p147258 --master-data=2 testdb -- CHANGE MASTER TO MASTER_LOG_FILE='mysql-bin.000011', MASTER_LOG_POS=107; --master-data=1时，备份文件中change master 是sql语句，在主从同步时用到 --master-data=2时，备份文件中--change master被注释掉，还原时不具有sql语句作用 例6：其它参数 --single-transaction 适合innodb事物数据库备份，原理是设定本次会话级别是repeatable read,保证本次会话备份时，不会看到其他已经提交的会话，保证数据一致性 -A, --all-databases Dump all the databases. This will be same as --databases -F, --flush-logs 即刷新binlog
总结：
innodb引擎 :一般生产全备使用的方法
[root@Darren2 tmp]# mysqldump -uroot -p147258 -A -B --master-data=2 --events --single-transaction > /tmp/all_$(date +%Y%m%d).bak
备份的全过程
下面mysqldump备份的过程：
root@localhost [testdb]>set global general_log=1;
[root@Darren1 data]# cat /dev/null > general.log
[root@Darren1 ~]# mysqldump -uroot -p147258 --single-transaction --master-data=2 testdb >testdb1_$(date +%Y%m%d)
[root@Darren1 data]# cat general.log
2016-12-21T15:31:00.474824Z 14 Connect root@localhost on using Socket
2016-12-21T15:31:00.475031Z 14 Query /*!40100 SET @@SQL_MODE='' */
2016-12-21T15:31:00.475194Z 14 Query /*!40103 SET TIME_ZONE='+00:00' */
2016-12-21T15:31:00.475282Z 14 Query FLUSH /*!40101 LOCAL */ TABLES --刷表，为了防止有表的DDL操作，如果备份的时候有表的DDL操作，flush tables一直处于等待状态，直到DDL动作结束，才执行flush tables
2016-12-21T15:31:00.475598Z 14 Query FLUSH TABLES WITH READ LOCK --全局锁表，所有会话不能对表进行任何DML和DDL操作，让数据处于一致性状态
2016-12-21T15:31:00.475661Z 14 Query SET SESSION TRANSACTION ISOLATION LEVEL REPEATABLE READ --设置隔离界别为RR
2016-12-21T15:31:00.475728Z 14 Query START TRANSACTION /*!40100 WITH CONSISTENT SNAPSHOT */ --开始事物
2016-12-21T15:31:00.475805Z 14 Query SHOW VARIABLES LIKE 'gtid\_mode' --查看GTID状态
2016-12-21T15:31:00.478393Z 14 Query SHOW MASTER STATUS --查看master
2016-12-21T15:31:00.478487Z 14 Query UNLOCK TABLES --解锁
2016-12-21T15:31:00.478625Z 14 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 ('testdb'))) GROUP BY LOGFILE_GROUP_NAME, FILE_NAME, ENGINE, TOTAL_EXTENTS, INITIAL_SIZE ORDER BY LOGFILE_GROUP_NAME
2016-12-21T15:31:00.480360Z 14 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 ('testdb')) ORDER BY TABLESPACE_NAME, LOGFILE_GROUP_NAME
2016-12-21T15:31:00.481072Z 14 Query SHOW VARIABLES LIKE 'ndbinfo\_version'
2016-12-21T15:31:00.483114Z 14 Init DB testdb
2016-12-21T15:31:00.483193Z 14 Query SAVEPOINT sp --创建事物的回滚点，如果下面一旦出错，可以回滚到回滚点之前的状态；
2016-12-21T15:31:00.483262Z 14 Query show tables
2016-12-21T15:31:00.483459Z 14 Query show table status like 't1'
2016-12-21T15:31:00.483711Z 14 Query SET SQL_QUOTE_SHOW_CREATE=1
2016-12-21T15:31:00.483782Z 14 Query SET SESSION character_set_results = 'binary'
2016-12-21T15:31:00.483844Z 14 Query show create table `t1`
2016-12-21T15:31:00.483927Z 14 Query SET SESSION character_set_results = 'utf8'
2016-12-21T15:31:00.483998Z 14 Query show fields from `t1`
2016-12-21T15:31:00.484307Z 14 Query show fields from `t1`
2016-12-21T15:31:00.484551Z 14 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `t1`
2016-12-21T15:31:00.484758Z 14 Query SET SESSION character_set_results = 'binary'
2016-12-21T15:31:00.484828Z 14 Query use `testdb`
2016-12-21T15:31:00.484888Z 14 Query select @@collation_database
2016-12-21T15:31:00.484962Z 14 Query SHOW TRIGGERS LIKE 't1'
2016-12-21T15:31:00.485199Z 14 Query SET SESSION character_set_results = 'utf8'
2016-12-21T15:31:00.485255Z 14 Query ROLLBACK TO SAVEPOINT sp --回到回滚点
2016-12-21T15:31:00.485315Z 14 Query show table status like 't2'
2016-12-21T15:31:00.485464Z 14 Query SET SQL_QUOTE_SHOW_CREATE=1
2016-12-21T15:31:00.485515Z 14 Query SET SESSION character_set_results = 'binary'
2016-12-21T15:31:00.485567Z 14 Query show create table `t2`
2016-12-21T15:31:00.485635Z 14 Query SET SESSION character_set_results = 'utf8'
2016-12-21T15:31:00.485744Z 14 Query show fields from `t2`
2016-12-21T15:31:00.485968Z 14 Query show fields from `t2`
2016-12-21T15:31:00.486185Z 14 Query SELECT /*!40001 SQL_NO_CACHE */ * FROM `t2`
2016-12-21T15:31:00.486298Z 14 Query SET SESSION character_set_results = 'binary'
2016-12-21T15:31:00.486351Z 14 Query use `testdb`
2016-12-21T15:31:00.486407Z 14 Query select @@collation_database
2016-12-21T15:31:00.486468Z 14 Query SHOW TRIGGERS LIKE 't2'
2016-12-21T15:31:00.486792Z 14 Query SET SESSION character_set_results = 'utf8'
2016-12-21T15:31:00.486887Z 14 Query ROLLBACK TO SAVEPOINT sp --回到回滚点
2016-12-21T15:31:00.486943Z 14 Query RELEASE SAVEPOINT sp --释放回滚点
2016-12-21T15:31:00.513639Z 14 Quit --退出
总结：mysqldump的过程：
flush tables; flush table with read lock; set tx_isolation='repeatable-read'; start transaction; GTID_MODE; show master stauts; unlock tables; SAVEPOINT sp show create table `t1` SELECT /*!40001 SQL_NO_CACHE */ * FROM `t1` SHOW TRIGGERS LIKE 't1' ROLLBACK TO SAVEPOINT sp RELEASE SAVEPOINT sp
注意事项：
在使用mysqldump备份表的时候，如果对备份的表进行DDL操作可能使备份失败，因为DDL不在事物的框架中，mysql8.0以后可能会把DDL放在事物框架中；
当有大事物更新写入的时候，flush logs，flush tables要等待或者卡顿,也有可能失败，这是常见的坑；