修改ibdata1大小的验证以及如何使用mysqld_multi管理多实例

Posted 2020-07-02 SlowTech

修改ibdata1大小的验证

ibdata是共享表空间，在mysql初始化的时候就生成了。

但很多童鞋会看到网上各种大神的调优建议，在MySQL已经初始化的情况下，修改配置文件中innodb_data_file_path=ibdata1:12M:autoextend。导致MySQL启动的时候报错。

下面来模拟一下：

原来的ibdata1大小为12M

# ll -h /sales3306/mysql/data/ibdata1 
-rw-rw----. 1 mysql mysql 12M Mar 31 21:09 /sales3306/mysql/data/ibdata1

修改配置文件中的innodb_data_file_path参数，调整其大小

innodb_data_file_path=ibdata1:20M:autoextend

重启数据库服务

启动数据库的过程中没有报错，但就是没有起来，查看日志信息

[ERROR] InnoDB: auto-extending data file /sales3306/mysql/data/ibdata1 is of a different size 768 pages (rounded down to MB) than specified in the .cnf file: initial 1280 pages, max 0 (relevant if non-zero) pages!

报错信息其实很明显，ibdata1实际大小和配置文件中指定的大小不一致。

遇到这种问题，如何修复呢？

其实，只需将该参数设置为等于或者小于其实际大小。

关于等于，其实不难理解，毕竟要吻合，那小于又为什么可以呢？关键在于该参数后面的autoextend选项，所以实际值比初始值大很正常。当然小于的情况只适用于带有autoextend选项的表空间，MySQL可指定多个表空间，但只有最后一个才能指定该选项。

获取其实际大小，有两种方式，

一是通过ls -l查看其具体大小，可直接写12582912（12M）

二是通过错误日志的报错信息，譬如上面很容易算出其实际大小为768*16/1024=12M

当然，如果默认的共享表空间体积太大了，可新增一个表空间

innodb_data_file_path=ibdata1:12M;ibdata2:20M:autoextend

注意：共享表空间，即便把数据清理掉后，也不会回收空间，只能迁移数据，重新初始化。

 

mysqld_multi的使用

1. 在执行mysqld_multi时报以下错误：

/usr/local/mysql/bin/mysqld_multi --defaults-file=/root/multi.cnf report
WARNING: my_print_defaults command not found.
Please make sure you have this command available and
in your path. The command is available from the latest
MySQL distribution.
ABORT: Can‘t find command ‘my_print_defaults‘.
This command is available from the latest MySQL
distribution. Please make sure you have the command
in your PATH.

解决方法：

在/etc/profile中添加如下内容：

export PATH=$PATH:/usr/local/mysql/bin/

并使其生效 source /etc/profile

2. 启动失败，通过查看mysqld_multi的错误日志/usr/local/mysql/multi.log 

有如下报错信息：

Starting MySQL servers

160116 20:25:22 mysqld_safe Logging to ‘/var/log/mariadb/mariadb.log‘.
touch: cannot touch ‘/var/log/mariadb/mariadb.log’: No such file or directory
chmod: cannot access ‘/var/log/mariadb/mariadb.log’: No such file or directory
touch: cannot touch ‘/var/log/mariadb/mariadb.log’: No such file or directory
chown: cannot access ‘/var/log/mariadb/mariadb.log’: No such file or directory
160116 20:25:22 mysqld_safe Logging to ‘/sales3307/mysql/log/.err‘.
160116 20:25:22 mysqld_safe Starting mysqld daemon with databases from /sales3306/mysql/data
/usr/local/mysql/bin/mysqld_safe: line 129: /var/log/mariadb/mariadb.log: No such file or directory
/usr/local/mysql/bin/mysqld_safe: line 166: /var/log/mariadb/mariadb.log: No such file or directory
touch: cannot touch ‘/var/log/mariadb/mariadb.log’: No such file or directory
chown: cannot access ‘/var/log/mariadb/mariadb.log’: No such file or directory
chmod: cannot access ‘/var/log/mariadb/mariadb.log’: No such file or directory
160116 20:25:22 mysqld_safe mysqld from pid file /var/run/mariadb/mariadb.pid ended
/usr/local/mysql/bin/mysqld_safe: line 129: /var/log/mariadb/mariadb.log: No such file or directory
160116 20:25:22 mysqld_safe Starting mysqld daemon with databases from /sales3307/mysql/data
160116 20:25:23 mysqld_safe mysqld from pid file /var/run/mariadb/mariadb.pid ended
mysqld_multi log file version 2.16; run: Sat Jan 16 20:25:24 2016

。。。怎么会出现mariadb的信息，这个跟我的操作系统有关，CentOS 7，默认的数据库是Mariadb，而不是MysQL。

失败原因：

没有指定错误日志

指定错误日志后，log-error=/sales3307/mysql/log/error.log

重新启动，又报如下错误：

2016-01-16 20:41:09 18683 [ERROR] /usr/local/mysql/bin/mysqld: Can‘t create/write to file ‘/var/run/mariadb/mariadb.pid‘ (Errcode: 2 - No such file or directory)
2016-01-16 20:41:09 18683 [ERROR] Can‘t start server: can‘t create PID file: No such file or directory
160116 20:41:10 mysqld_safe mysqld from pid file /var/run/mariadb/mariadb.pid ended

失败原因：没有指定pid文件

指定pid文件的路径

pid-file=/sales3307/mysql/run/mysqld.pid

重新启动，终于启动成功

# mysqld_multi --defaults-file=/root/multi.cnf report
Reporting MySQL servers
MySQL server from group: mysqld3306 is running
MySQL server from group: mysqld3307 is running

最后贴一下配置文件：

mysqld的部分没有贴，这块配置是公用的

[mysqld_multi]
mysqld     = /usr/local/mysql/bin/mysqld_safe
mysqladmin = /usr/local/mysql/bin/mysqladmin
#用来做关闭mysql使用
user = root
log=/usr/local/mysql/multi.log
##password=‘‘

[mysqld3306]
port=3306
socket=/sales3306/mysql/run/mysql.sock
datadir=/sales3306/mysql/data
server-id=1003306
log-bin=/sales3306/mysql/log/mysql-bin
tmpdir=/sales3306/mysql/tmp/
innodb_log_group_home_dir = /sales3306/mysql/data
innodb_buffer_pool_size=200M
log-error=/sales3306/mysql/log/error.log
pid-file=/sales3306/mysql/run/mysqld.pid

[mysqld3307]
port=3307
socket=/sales3307/mysql/run/mysql.sock
datadir=/sales3307/mysql/data
server-id=1003307
log-bin=/sales3307/mysql/log/mysql-bin
tmpdir=/sales3307/mysql/tmp/
innodb_log_group_home_dir = /sales3307/mysql/data
innodb_buffer_pool_size=100M
log-error=/sales3307/mysql/log/error.log
pid-file=/sales3307/mysql/run/mysqld.pid

 

在折腾过程中还是蛮多坎坷的，结论就是很个性化的定制，譬如socket，log-error，pid-file都要指定各自的路径。不然启动过程中，系统会按默认的来，多实例都按默认的来，会起冲突的。但整个排错过程还是蛮简单的，不是看mysqld_multi的错误日志，就是实例本身的日志。

 

下面演示一下，mysqld_multi的用法：

[[email protected] ~]# mysqld_multi --defaults-file=/root/multi.cnf stop

[[email protected] ~]# mysqld_multi --defaults-file=/root/multi.cnf report
Reporting MySQL servers
MySQL server from group: mysqld3306 is not running
MySQL server from group: mysqld3307 is not running

[[email protected] ~]# mysqld_multi --defaults-file=/root/multi.cnf report 3306
Reporting MySQL servers
MySQL server from group: mysqld3306 is not running

[[email protected] ~]# mysqld_multi --defaults-file=/root/multi.cnf start 3306


[[email protected] ~]# mysqld_multi --defaults-file=/root/multi.cnf report
Reporting MySQL servers
MySQL server from group: mysqld3306 is running
MySQL server from group: mysqld3307 is not running

在上述[mysqld_multi]的配置中，有个user和password，这个是用来执行mysqladmin关闭程序的。

有时候，用root权限太大，且密码以明文的方式暴露存在安全风险。

虽然mysqld_multi支持--password选项，但如果两个实例的密码不一样，又如何同时关闭实例呢？

可为两个实例创建同名账户，只赋予shutdown权限，这样可解决上述的困扰。

mysql>  grant shutdown on *.* to ‘multiadmin‘@‘localhost‘ identified by ‘123456‘;
Query OK, 0 rows affected (0.00 sec)

mysql> flush privileges;
Query OK, 0 rows affected (0.00 sec)