mysql 多主一从，分表同步.可否

Posted 2023-04-12

参考技术A 据我所知应该是做不到的,主从设置只能到某个数据库,而不能到某个表里的记录。 只能通过其他方式进行,比如自己写shell脚本

多主一从mysql replication同步表的大胆尝试.

能否将不同机器上的不同库中的表同步到同一个机器的同一个库中？表是不同的.而且对于slave这台机子来说,这些表只用来读.

 

同步不同库的表很简单了,用

replicate-do-table=db_name.tbl_name

多个不同机子上的不同库的表同步到同一个机子的同一个库中,再加上下面这句.

replicate-rewrite-db=from_name->to_name

配置slave 上的my.cnf

cat my.cnf

[mysqld_multi]
mysqld = /usr/local/mysql/bin/mysqld_safe
mysqladmin = /usr/local/mysql/bin/mysqladmin

[mysqld1]
port     = 3306
socket   = /tmp/mysql.sock1
skip-locking
pid-file=/usr/local/mysql/data/test-db1a.pid
datadir = /usr/local/mysql/data

log=/usr/local/mysql/data/test-db1.log
user = mysql
log-slow-queries=/usr/local/mysql/data/slowquery1.log
long_query_time = 2
key_buffer = 256M
max_allowed_packet = 1M
table_cache = 512
sort_buffer_size = 2M
read_buffer_size = 2M
myisam_sort_buffer_size = 64M
thread_cache = 32
query_cache_size = 32M
thread_concurrency = 2
max_connections=500
server-id     = 2
master-host   =   192.168.0.1
master-user   =   ‘replcate‘
master-password =   ‘1234567890‘
master-port   = 3306
report-host = test-db3
master-connect-retry = 30
log-bin
log-slave-updates

replicate-rewrite-db=from_name->to_name

replicate-do-table=db_name.tbl_name
[mysqld2]
port = 3307
socket = /tmp/mysql.sock2
pid-file = /usr/local/mysql/data/test-db2b.pid
datadir = /usr/local/mysql/data
log=/usr/local/mysql/data/test-db2.log
user = mysql
log-slow-queries=/usr/local/mysql/data/slowquery2.log
long_query_time = 10
key_buffer = 128M
max_allowed_packet = 1M
table_cache = 512
sort_buffer_size = 1M
read_buffer_size = 1M
myisam_sort_buffer_size = 32M
thread_cache = 32
query_cache_size = 16M
thread_concurrency = 2
max_connections=300
server-id     = 2
master-host   =   192.168.0.2
master-user   =   ‘repl‘
master-password =   ‘1234567890‘
master-port   = 3307
report-host = test-db3
master-connect-retry = 30
log-bin
log-slave-updates

replicate-rewrite-db=from_name->to_name

replicate-do-table=db_name.tbl_name

好了,看看上面配置,明白了两个mysqld实例用了相同的库表文件,会不会有问题呢?

当然,使用不当的话会有问题,首先要符合两个实例读写的对象没有交叉.(同步过来的表唯一),其次选择合适的存储引擎,MYISM比innodb从理论上更适合这种应用．再次使用过程中对slave上的两个mysqld实例，只使用一个进行本机其它库表（非同步过来的）写操作．

ok,这种解决方案已通过测试．

 

来源： http://blog.chinaunix.net/uid-220350-id-149202.html