读写分离，简单的说就是把对数据库的读与写操作分开，以对应不同的数据库服务器。主数据库提供写操作，从数据库提供读操作，这样能有效地减轻单台数据库的压力。在很多系统中读操作的比例远高于写操作，所以对应的读操作的数据库可以有多台，通过负载均衡技术进一步分摊读操作的压力，让整个数据库系统高效，平稳地运行。而这些读写分离所需要的复杂数据库架构，要对开发人员和应用程序透明。这里我们通过mycat来实现上述的功能。
mycat读写分离的控制由mycat的schema.xml配置文件中datahost标签的balance属性来实现。
基于mysql的读写分离
首先搭建MySQL的主从分离，因为这里只是测试mysql的读写分离，不再说明主从复制的搭建过程。主从搭建。
主服务器：10.0.102.221 从服务器：10.0.102.222
备主服务器：10.0.102.179
搭建过程： 在221和222之间搭建主从架构，在221和179之间搭建互为主从架构即可。也就是这三台服务器构成一个双主一从架构，主意和MySQL5.7的多源复制架构区分。
多源复制
搭建好测试需要的MySQL架构之后，然后开始配置mycat的读写分离。schema.xml中的主要部分配置文件如下：
<schema name="TESTDB" checkSQLschema="false" sqlMaxLimit="100"> <!-- auto sharding by id (long) <table name="employee" dataNode="dn1,dn2,dn3" rule="sharding-by-murmur" /> --> <table name="employee" dataNode="dn1"/> #在这里没有分片操作，因此没有rule属性 </schema> <!-- <dataNode name="dn1$0-743" dataHost="localhost1" database="db$0-743" /> --> <dataNode name="dn1" dataHost="split-one" database="mytest" /> #指定dataNode <!--<dataNode name="dn4" dataHost="sequoiadb1" database="SAMPLE" /> <dataNode name="jdbc_dn3" dataHost="jdbchost" database="db3" /> --> <dataHost name="split-one" maxCon="1000" minCon="10" balance="0" writeType="0" dbType="mysql" dbDriver="native" switchType="1" slaveThreshold="100"> <heartbeat>select user()</heartbeat>
#这个writeHost标签包含的一个主从的架构配置，w为master服务，r为slave服务器 <writeHost host="hostM1" url="10.0.102.221:3306" user="root" password="123456"> <readHost host="hostS2" url="10.0.102.222:3306" user="root" password="123456" /> </writeHost>
#这个writehost标签包含的是一个备用的主服务器。 <writeHost host="hostM2" url="10.0.102.179:3306" user="root" password="123456" /> </dataHost>
上面配置修改成功之后，启动mycat，登录到管理端口：
mysql> show @@datasource; +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | DATANODE | NAME | TYPE | HOST | PORT | W/R | ACTIVE | IDLE | SIZE | EXECUTE | READ_LOAD | WRITE_LOAD | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | dn1 | hostM1 | mysql | 10.0.102.221 | 3306 | W | 0 | 10 | 1000 | 14 | 0 | 0 | | dn1 | hostM2 | mysql | 10.0.102.179 | 3306 | W | 0 | 1 | 1000 | 4 | 0 | 0 | | dn1 | hostS2 | mysql | 10.0.102.222 | 3306 | R | 0 | 3 | 1000 | 7 | 0 | 0 | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ 3 rows in set (0.01 sec)
上面提到过，mycat的主从分离是有参数balance控制,下面根据不同的balance值测试对应读写分离。
balance值为0时：不开启读写分离机制，所有的读操作都发送到当前可用的writeHost上。
#插入数据，写入数据，发现这时候所有的读写都分配到当前可用的writeHost上，也就是主master上。
mysql> show @@datasource; +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | DATANODE | NAME | TYPE | HOST | PORT | W/R | ACTIVE | IDLE | SIZE | EXECUTE | READ_LOAD | WRITE_LOAD | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | dn1 | hostM1 | mysql | 10.0.102.221 | 3306 | W | 0 | 10 | 1000 | 94 | 11 | 1 | | dn1 | hostM2 | mysql | 10.0.102.179 | 3306 | W | 0 | 1 | 1000 | 72 | 0 | 0 | | dn1 | hostS2 | mysql | 10.0.102.222 | 3306 | R | 0 | 6 | 1000 | 78 | 0 | 0 | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ 3 rows in set (0.00 sec) mysql>
balance值为1时：全部的readHost与stand by writeHost都参与select语句的负载均衡。
mysql> show @@datasource; +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | DATANODE | NAME | TYPE | HOST | PORT | W/R | ACTIVE | IDLE | SIZE | EXECUTE | READ_LOAD | WRITE_LOAD | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | dn1 | hostM1 | mysql | 10.0.102.221 | 3306 | W | 0 | 10 | 1000 | 19 | 0 | 1 | | dn1 | hostM2 | mysql | 10.0.102.179 | 3306 | W | 0 | 1 | 1000 | 18 | 10 | 0 | | dn1 | hostS2 | mysql | 10.0.102.222 | 3306 | R | 0 | 3 | 1000 | 16 | 5 | 0 | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ 3 rows in set (0.00 sec)
#插入一条数据，是写在hostM1上，读的操作则分布在hostM2和hostS2上，测试select查询时，可以多查询几次。
当balance值为2时： 所有的读操作都随机地在writeHost, readHost上分发。
mysql> show @@datasource; +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | DATANODE | NAME | TYPE | HOST | PORT | W/R | ACTIVE | IDLE | SIZE | EXECUTE | READ_LOAD | WRITE_LOAD | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | dn1 | hostM1 | mysql | 10.0.102.221 | 3306 | W | 0 | 10 | 1000 | 20 | 6 | 0 | | dn1 | hostM2 | mysql | 10.0.102.179 | 3306 | W | 0 | 1 | 1000 | 8 | 4 | 0 | | dn1 | hostS2 | mysql | 10.0.102.222 | 3306 | R | 0 | 3 | 1000 | 14 | 7 | 0 | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ 3 rows in set (0.00 sec)
#整个架构中所有的服务器都参与读的均衡，然后再插入数据，发现还是只有主master上写！
mysql> show @@datasource;
+----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+
| DATANODE | NAME | TYPE | HOST | PORT | W/R | ACTIVE | IDLE | SIZE | EXECUTE | READ_LOAD | WRITE_LOAD |
+----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+
| dn1 | hostM1 | mysql | 10.0.102.221 | 3306 | W | 0 | 10 | 1000 | 32 | 6 | 2 |
| dn1 | hostM2 | mysql | 10.0.102.179 | 3306 | W | 0 | 1 | 1000 | 18 | 4 | 0 |
| dn1 | hostS2 | mysql | 10.0.102.222 | 3306 | R | 0 | 3 | 1000 | 24 | 7 | 0 |
+----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+
3 rows in set (0.01 sec)
当balance值为3时： 所有的读请求都随机分配到writeHost对应的readHost上执行，writeHost不负担读压力，注意balance值为3只在mycat1.4版本及之后的版本中有。
mysql> show @@datasource; +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | DATANODE | NAME | TYPE | HOST | PORT | W/R | ACTIVE | IDLE | SIZE | EXECUTE | READ_LOAD | WRITE_LOAD | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | dn1 | hostM1 | mysql | 10.0.102.221 | 3306 | W | 0 | 10 | 1000 | 14 | 0 | 0 | | dn1 | hostM2 | mysql | 10.0.102.179 | 3306 | W | 0 | 1 | 1000 | 4 | 0 | 0 | | dn1 | hostS2 | mysql | 10.0.102.222 | 3306 | R | 0 | 3 | 1000 | 17 | 10 | 0 | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ 3 rows in set (0.00 sec)
#所有的查询都被分配到对应的slave上面。
基于mariadb的galera集群做读写分离
mariadb的galera集群时真正的多主架构，集群中的每台服务器都可以写入数据。首先搭建mariadb集群，过程略；搭建之后，配置mycat的读写分离。
schema.xml配置文件如下，和之前的mysql主从架构配置会有一些不同。
<schema name="TESTDB" checkSQLschema="false" sqlMaxLimit="100"> <!-- auto sharding by id (long) <table name="employee" dataNode="dn1,dn2,dn3" rule="sharding-by-murmur" /> --> <table name="employee" dataNode="dn1"/> </schema> <!-- <dataNode name="dn1$0-743" dataHost="localhost1" database="db$0-743" /> --> <dataNode name="dn1" dataHost="split-one" database="mytest" /> <!--<dataNode name="dn4" dataHost="sequoiadb1" database="SAMPLE" /> <dataNode name="jdbc_dn3" dataHost="jdbchost" database="db3" /> --> <dataHost name="split-one" maxCon="1000" minCon="10" balance="0" writeType="0" dbType="mysql" dbDriver="native" switchType="1" slaveThreshold="100"> <heartbeat>show status like "wsrep%"</heartbeat> <writeHost host="hostM1" url="10.0.102.204:3306" user="root" password="123456"> </writeHost> <writeHost host="hostM2" url="10.0.102.221:3306" user="root" password="123456" > </writeHost> <writeHost host="hostM3" url="10.0.102.222:3306" user="root" password="123456" > </writeHost> </dataHost>
mysql> show @@datasource; +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | DATANODE | NAME | TYPE | HOST | PORT | W/R | ACTIVE | IDLE | SIZE | EXECUTE | READ_LOAD | WRITE_LOAD | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ | dn1 | hostM1 | mysql | 10.0.102.204 | 3306 | W | 0 | 1 | 1000 | 1 | 0 | 0 | | dn1 | hostM2 | mysql | 10.0.102.221 | 3306 | W | 0 | 10 | 1000 | 11 | 0 | 0 | | dn1 | hostM3 | mysql | 10.0.102.222 | 3306 | W | 0 | 1 | 1000 | 1 | 0 | 0 | +----------+--------+-------+--------------+------+------+--------+------+------+---------+-----------+------------+ 3 rows in set (0.01 sec) mysql> show @@heartbeat; +--------+-------+--------------+------+---------+-------+--------+---------+--------------+---------------------+-------+ | NAME | TYPE | HOST | PORT | RS_CODE | RETRY | STATUS | TIMEOUT | EXECUTE_TIME | LAST_ACTIVE_TIME | STOP | +--------+-------+--------------+------+---------+-------+--------+---------+--------------+---------------------+-------+ | hostM1 | mysql | 10.0.102.204 | 3306 | 1 | 0 | idle | 0 | 19,19,19 | 2019-01-21 14:43:51 | false | | hostM2 | mysql | 10.0.102.221 | 3306 | 1 | 0 | idle | 0 | 16,16,16 | 2019-01-21 14:43:51 | false | | hostM3 | mysql | 10.0.102.222 | 3306 | 1 | 0 | idle | 0 | 17,17,17 | 2019-01-21 14:43:51 | false | +--------+-------+--------------+------+---------+-------+--------+---------+--------------+---------------------+-------+ 3 rows in set (0.00 sec) mysql>
mariadb的读写分离基本与上面mysql主从保持一致，不再重复测试。
当balance的值为0时：不进行读写分离，所有的读写都走配置的dataHost的第一个writeHost。
当balance值为1时：全部的readHost和stand by writeHost都参与select语句的负载均衡。如果第一个节点未带slave节点，则三个节点中，一个作为写，另外两个作为读
当balance为2时，所有的读操作随机地在writeHost，readHost上分发。
balance值为3时，所有的读请求都随机地分发至writeHost对应的readHost执行，writeHost不负担读的压力。如果writeHost没有对应的readHost，则查询分发至writeHost。
测试的过程中需要注意：一条insert语句插入多条记录，write_load的值只增加1.
在插入数据测试的时候，可以采用如下方式查看，插入一个特殊的值，来确定数据究竟插入到了哪一个服务器中。
mysql> insert into employee(id, name) values(4, @@HOSTNAME); Query OK, 1 row affected (0.04 sec) mysql> select * from employee where id =4; +------+-------+ | id | name | +------+-------+ | 4 | test2 | +------+-------+ 1 row in set (0.06 sec) mysql>