MyCAT部署及实现读写分离(转)
MyCAT是mysql中间件,前身是阿里大名鼎鼎的Cobar,Cobar在开源了一段时间后,不了了之。于是MyCAT扛起了这面大旗,在大数据时代,其重要性愈发彰显。这篇文章主要是MyCAT的入门部署。
一、安装java
因Mycat是用java开发的,所以需要在实验环境下安装java,官方建议jdk1.7及以上版本
Java Oracle官方下载地址为:
http://www.oracle.com/technetwork/java/javase/archive-139210.html
解压jdk-7u79-linux-x64.tar.gz文件,并配置java的环境变量
# tar xvf jdk-7u79-linux-x64.tar.gz
# mv jdk1.7.0_79/ /usr/local/
编辑/etc/profile文件
# vim /etc/profile 添加如下内容
export JAVA_HOME=/usr/local/jdk1.7.0_79 export PATH=$JAVA_HOME/bin:$PATH export CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar
# source /etc/profile --使/etc/profile文件生效
# java -version
java version "1.7.0_79" Java(TM) SE Runtime Environment (build 1.7.0_79-b15) Java HotSpot(TM) 64-Bit Server VM (build 24.79-b02, mixed mode)
二、安装MySQL
MySQL官方下载地址如下:
http://dev.mysql.com/downloads/mysql/
里面有各种版本,包括RPM,二进制,源码包。
为了方便起见,我这里选择是RPM包
# yum install -y mysql-community-server-5.6.26-2.el5.x86_64.rpm
启动MySQL
# /etc/init.d/mysqld start
Starting mysqld (via systemctl): [ OK ]
为root账户创建密码
# mysqladmin -u root password "123456"
建议为123456,后面MyCAT配置文件schema.xml中会用到
三、安装MyCAT
创建MyCAT用户并设置密码
# useradd mycat
# passwd mycat
解压MyCAT文件
# tar xvf Mycat-server-1.5-alpha-20151221110028-linux.tar.gz
# mv mycat/ /usr/local/
设置mycat目录的属主和属组
# chown -R mycat.mycat /usr/local/mycat/
四、测试MyCAT
首先在MySQL上创建三个数据库:db1,db2,db3。
mysql> create database db1; Query OK, 1 row affected (0.00 sec) mysql> create database db2; Query OK, 1 row affected (0.00 sec) mysql> create database db3; Query OK, 1 row affected (0.00 sec)
启动mycat服务
# cd /usr/local/mycat/bin/
# ./mycat start
Starting Mycat-server...
查看mycat服务是否启动
# ps -ef |grep mycat
root 9640 7257 0 22:47 pts/3 00:00:00 grep --color=auto mycat
并没有启动
查看日志
# cd /usr/local/mycat/logs/
# cat wrapper.log
STATUS | wrapper | 2016/01/07 22:44:23 | --> Wrapper Started as Daemon STATUS | wrapper | 2016/01/07 22:44:23 | Launching a JVM... ERROR | wrapper | 2016/01/07 22:44:25 | JVM exited while loading the application. INFO | jvm 1 | 2016/01/07 22:44:25 | Error: Exception thrown by the agent : java.net.MalformedURLException: Local host name unknown: java.net.UnknownHostException: MySQL-Server1: MySQL-Server1: Name or service not known STATUS | wrapper | 2016/01/07 22:44:29 | Launching a JVM... ERROR | wrapper | 2016/01/07 22:44:29 | JVM exited while loading the application. INFO | jvm 2 | 2016/01/07 22:44:29 | Error: Exception thrown by the agent : java.net.MalformedURLException: Local host name unknown: java.net.UnknownHostException: MySQL-Server1: MySQL-Server1: Name or service not known STATUS | wrapper | 2016/01/07 22:44:34 | Launching a JVM... ERROR | wrapper | 2016/01/07 22:44:34 | JVM exited while loading the application. INFO | jvm 3 | 2016/01/07 22:44:34 | Error: Exception thrown by the agent : java.net.MalformedURLException: Local host name unknown: java.net.UnknownHostException: MySQL-Server1: MySQL-Server1: Name or service not known STATUS | wrapper | 2016/01/07 22:44:38 | Launching a JVM... ERROR | wrapper | 2016/01/07 22:44:38 | JVM exited while loading the application. INFO | jvm 4 | 2016/01/07 22:44:38 | Error: Exception thrown by the agent : java.net.MalformedURLException: Local host name unknown: java.net.UnknownHostException: MySQL-Server1: MySQL-Server1: Name or service not known STATUS | wrapper | 2016/01/07 22:44:42 | Launching a JVM... ERROR | wrapper | 2016/01/07 22:44:43 | JVM exited while loading the application. INFO | jvm 5 | 2016/01/07 22:44:43 | Error: Exception thrown by the agent : java.net.MalformedURLException: Local host name unknown: java.net.UnknownHostException: MySQL-Server1: MySQL-Server1: Name or service not known FATAL | wrapper | 2016/01/07 22:44:43 | There were 5 failed launches in a row, each lasting less than 300 seconds. Giving up. FATAL | wrapper | 2016/01/07 22:44:43 | There may be a configuration problem: please check the logs. STATUS | wrapper | 2016/01/07 22:44:43 | <-- Wrapper Stopped
根据报错信息,怀疑主机名没有绑定
修改hosts文件,绑定主机名
# vim /etc/hosts
127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4 ::1 localhost localhost.localdomain localhost6 localhost6.localdomain6 192.168.244.144 MySQL-Server1
重新启动mycat服务
# cd /usr/local/mycat/bin/
# ./mycat start
这次mycat正常启动
# ps -ef |grep mycat
root 10725 1 0 22:54 ? 00:00:00 /usr/local/mycat/bin/./wrapper-linux-x86-64 /usr/local/mycat/conf/wrapper.conf wrapper.syslog.ident=mycat wrapper.pidfile=/usr/local/mycat/logs/mycat.pid wrapper.daemonize=TRUE wrapper.lockfile=/var/lock/subsys/mycat root 10881 7257 0 22:55 pts/3 00:00:00 grep --color=auto mycat
下面以Travelrecord为例,来进行插入,查询,路由分析等基本操作。
首先利用mycat连接mysql数据库
# mysql -utest -ptest -h127.0.0.1 -P8066 -DTESTDB
其中8066是mycat的监听端口,类似于mysql的3306端口,其中-u,-p,-h分别是用户名,密码和主机,-D是连接的逻辑库。
至于为什么是这些,这个跟配置文件有关。
红色部分表明连接的是mycat。
创建Travelrecord表
create table travelrecord (id bigint not null primary key,user_id varchar(100),traveldate DATE, fee decimal,days int);
插入数据
mysql> insert into travelrecord(id,user_id,traveldate,fee,days) values(1,'Victor',20160101,100,10); Query OK, 1 row affected (0.00 sec) mysql> insert into travelrecord(id,user_id,traveldate,fee,days) values(5000001,'Job',20160102,100,10); Query OK, 1 row affected (0.00 sec) mysql> insert into travelrecord(id,user_id,traveldate,fee,days) values(10000001,'Slow',20160103,100,10); Query OK, 1 row affected (0.00 sec)
至于ID为什么取三个值,这个与conf目录下autopartition-long.txt的定义有关,这个文件主要定义auto-sharding-long的规则。
# range start-end ,data node index # K=1000,M=10000. 0-500M=0 500M-1000M=1 1000M-1500M=2
我主要是测试在id取不同区间的值时,分片的效果。
下面来看看分片的效果
mysql> select * from db1.travelrecord; +----+---------+------------+------+------+ | id | user_id | traveldate | fee | days | +----+---------+------------+------+------+ | 1 | Victor | 2016-01-01 | 100 | 10 | +----+---------+------------+------+------+ 1 row in set (0.00 sec) mysql> select * from db2.travelrecord; +---------+---------+------------+------+------+ | id | user_id | traveldate | fee | days | +---------+---------+------------+------+------+ | 5000001 | Job | 2016-01-02 | 100 | 10 | +---------+---------+------------+------+------+ 1 row in set (0.00 sec) mysql> select * from db3.travelrecord; +----------+---------+------------+------+------+ | id | user_id | traveldate | fee | days | +----------+---------+------------+------+------+ | 10000001 | Slow | 2016-01-03 | 100 | 10 | +----------+---------+------------+------+------+ 1 row in set (0.00 sec)
果然是分配到三个不同的节点上了。
如果想看MyCAT具体会将数据分配到哪个节点上,可通过路由分析。
语法其实蛮简单,就是SQL语句前加上explain语句。
下面根据explain命令查看create语句和insert语句具体会分配到哪些Datanode上。
mysql> explain create table travelrecord (id bigint not null primary key,user_id varchar(100),traveldate DATE, fee decimal,days int); +-----------+-----------------------------------------------------------------------------------------------------------------------+ | DATA_NODE | SQL | +-----------+-----------------------------------------------------------------------------------------------------------------------+ | dn1 | create table travelrecord (id bigint not null primary key,user_id varchar(100),traveldate DATE, fee decimal,days int) | | dn2 | create table travelrecord (id bigint not null primary key,user_id varchar(100),traveldate DATE, fee decimal,days int) | | dn3 | create table travelrecord (id bigint not null primary key,user_id varchar(100),traveldate DATE, fee decimal,days int) | +-----------+-----------------------------------------------------------------------------------------------------------------------+ 3 rows in set (0.00 sec) mysql> explain insert into travelrecord(id,user_id,traveldate,fee,days) values(1,'Victor',20160101,100,10); +-----------+----------------------------------------------------------------------------------------------+ | DATA_NODE | SQL | +-----------+----------------------------------------------------------------------------------------------+ | dn1 | insert into travelrecord(id,user_id,traveldate,fee,days) values(1,'Victor',20160101,100,10) | +-----------+----------------------------------------------------------------------------------------------+ 1 row in set (0.01 sec) mysql> explain insert into travelrecord(id,user_id,traveldate,fee,days) values(5000001,'Job',20160102,100,10); +-----------+-------------------------------------------------------------------------------------------------+ | DATA_NODE | SQL | +-----------+-------------------------------------------------------------------------------------------------+ | dn2 | insert into travelrecord(id,user_id,traveldate,fee,days) values(5000001,'Job',20160102,100,10) | +-----------+-------------------------------------------------------------------------------------------------+ 1 row in set (0.00 sec) mysql> explain insert into travelrecord(id,user_id,traveldate,fee,days) values(10000001,'Slow',20160103,100,10); +-----------+---------------------------------------------------------------------------------------------------+ | DATA_NODE | SQL | +-----------+---------------------------------------------------------------------------------------------------+ | dn3 | insert into travelrecord(id,user_id,traveldate,fee,days) values(10000001,'Slow',20160103,100,10) | +-----------+---------------------------------------------------------------------------------------------------+ 1 row in set (0.00 sec)
总结:
关于MyCAT的配置其实是蛮简单的,最主要的是熟悉各配置文件的规则。以上用户名,密码,如何分库,都是在配置文件中定义的,后续,有时间再一一详表。
关于配置文件,conf目录下主要以下三个需要熟悉。
server.xml是Mycat服务器参数调整和用户授权的配置文件
schema.xml是逻辑库定义和表以及分片定义的配置文件
rule.xml是分片规则的配置文件
在MySQL中间件出现之前,对于MySQL主从集群,如果要实现其读写分离,一般是在程序端实现,这样就带来一个问题,即数据库和程序的耦合度太高,如果我数据库的地址发生改变了,那么我程序端也要进行相应的修改,如果数据库不小心挂掉了,则同时也意味着程序的不可用,而这对很多应用来说,并不能接受。
引入MySQL中间件能很好的对程序端和数据库进行解耦,这样,程序端只需关注数据库中间件的地址,而无需知晓底层数据库是如何提供服务。
作为当前炙手可热的MySQL中间件,MyCAT实现MySQL主从集群的读写分离自是应有之义,其配置也相当简单。
在这里,我用三个实例组成MySQL主从集群,来验证MyCAT的读写分离功能,其实,一主一从就可以满足,之所以用三个,是为了验证MyCAT的分片功能。
集群组成如下:
角色 主机名 主机IP
master mysql-server1 192.168.244.145
slave mysql-server2 192.168.244.146
slave mysql-server3 192.168.244.144
在这里,还是使用Travelrecord表进行测试。
首先编辑MyCAT的配置文件schema.xml,关于dataHost的配置信息如下:
<dataHost name="localhost1" maxCon="1000" minCon="10" balance="1" writeType="0" dbType="mysql" dbDriver="native" switchType="-1" slaveThreshold="100"> <heartbeat>select user()</heartbeat> <!-- can have multi write hosts --> <writeHost host="hostM1" url="localhost:3306" user="root" password="123456"> </writeHost> <writeHost host="hostS1" url="192.168.244.146:3306" user="root" password="123456" /> <writeHost host="hostS2" url="192.168.244.144:3306" user="root" password="123456" /> </dataHost>
这里面,有两个参数需要注意,balance和 switchType。
其中,balance指的负载均衡类型,目前的取值有4种:
1. balance="0", 不开启读写分离机制,所有读操作都发送到当前可用的writeHost上。
2. balance="1",全部的readHost与stand by writeHost参与select语句的负载均衡,简单的说,当双主双从模式(M1->S1,M2->S2,并且M1与 M2互为主备),正常情况下,M2,S1,S2都参与select语句的负载均衡。
3. balance="2",所有读操作都随机的在writeHost、readhost上分发。
4. balance="3",所有读请求随机的分发到wiriterHost对应的readhost执行,writerHost不负担读压力
switchType指的是切换的模式,目前的取值也有4种:
1. switchType='-1' 表示不自动切换
2. switchType='1' 默认值,表示自动切换
3. switchType='2' 基于MySQL主从同步的状态决定是否切换,心跳语句为 show slave status
4. switchType='3'基于MySQL galary cluster的切换机制(适合集群)(1.4.1),心跳语句为 show status like 'wsrep%'。
因此,该配置文件中的balance="1"意味着作为stand by writeHost的hostS1和hostS2将参与select语句的负载均衡,这就实现了主从的读写分离,switchType='-1'意味着当主挂掉的时候,不进行自动切换,即hostS1和hostS2并不会被提升为主,仍只提供读的功能。这就避免了将数据读进slave的可能性,毕竟,单纯的MySQL主从集群并不允许将数据读进slave中,除非配置的是双master。
验证读写分离
下面来验证一下,
创建Travelrecord表
create table travelrecord (id bigint not null primary key,user_id varchar(100),traveldate DATE, fee decimal,days int);
插入数据
mysql> insert into travelrecord(id,user_id,traveldate,fee,days) values(1,@@hostname,20160101,100,10); Query OK, 1 row affected, 1 warning (0.02 sec) mysql> insert into travelrecord(id,user_id,traveldate,fee,days) values(5000001,@@hostname,20160102,100,10); Query OK, 1 row affected, 1 warning (0.01 sec)
在这里,用了一个取巧的方法,即对user_id插入了当前实例的主机名,这样可直观的观察读写是否分离以及MyCAT的分片功能。能这样做的原因在于我当前的MySQL版本-5.6.26默认是基于statement的复制,如果是基于row的复制,则这个方法将不可取。
查询数据
mysql> select * from travelrecord; +---------+---------------+------------+------+------+ | id | user_id | traveldate | fee | days | +---------+---------------+------------+------+------+ | 1 | mysql-server2 | 2016-01-01 | 100 | 10 | | 5000001 | mysql-server3 | 2016-01-02 | 100 | 10 | +---------+---------------+------------+------+------+ 2 rows in set (0.01 sec) mysql> select * from travelrecord; +---------+---------------+------------+------+------+ | id | user_id | traveldate | fee | days | +---------+---------------+------------+------+------+ | 5000001 | mysql-server3 | 2016-01-02 | 100 | 10 | | 1 | mysql-server2 | 2016-01-01 | 100 | 10 | +---------+---------------+------------+------+------+ 2 rows in set (0.02 sec) mysql> select * from travelrecord; +---------+---------------+------------+------+------+ | id | user_id | traveldate | fee | days | +---------+---------------+------------+------+------+ | 5000001 | mysql-server3 | 2016-01-02 | 100 | 10 | | 1 | mysql-server3 | 2016-01-01 | 100 | 10 | +---------+---------------+------------+------+------+ 2 rows in set (0.01 sec) mysql> select * from travelrecord; +---------+---------------+------------+------+------+ | id | user_id | traveldate | fee | days | +---------+---------------+------------+------+------+ | 5000001 | mysql-server3 | 2016-01-02 | 100 | 10 | | 1 | mysql-server3 | 2016-01-01 | 100 | 10 | +---------+---------------+------------+------+------+ 2 rows in set (0.01 sec) mysql> select * from travelrecord; +---------+---------------+------------+------+------+ | id | user_id | traveldate | fee | days | +---------+---------------+------------+------+------+ | 1 | mysql-server2 | 2016-01-01 | 100 | 10 | | 5000001 | mysql-server2 | 2016-01-02 | 100 | 10 | +---------+---------------+------------+------+------+
从上面的输出结果,可以得出以下两点:
一、该配置已实现读写分离,读出来的数据没有master节点的。
二、MyCAT的随机分发不是基于statement的,即一个select语句查询其中一个节点,另外一个select语句查询另外一个节点。它分发针对的是片的,同一个select语句的结果是有不同dataNode返回的。
不仅如此,从MyCAT日志中也可以获取读写分离的相关信息,当然,前提是MyCAT的日志级别是debug。日志相关信息如下:
验证mater挂了,slave还能提供读的功能
对于MySQL主从集群,我们的需求是master挂了,slave还能提供读的功能。
下面来测试一下
首先,人为的关闭主库
[root@mysql-server1 ~]# /etc/init.d/mysqld stop
登录MyCAT
[root@mysql-server1 ~]# mysql -utest -ptest -h127.0.0.1 -P8066 -DTESTDB
插入数据
mysql> insert into travelrecord(id,user_id,traveldate,fee,days) values(10000001,@@hostname,20160103,100,10); ERROR 1184 (HY000): Connection refused mysql> select * from travelrecord; +---------+---------------+------------+------+------+ | id | user_id | traveldate | fee | days | +---------+---------------+------------+------+------+ | 1 | mysql-server2 | 2016-01-01 | 100 | 10 | | 5000001 | mysql-server3 | 2016-01-02 | 100 | 10 | +---------+---------------+------------+------+------+ 2 rows in set (0.02 sec)
可见无法插入数据,但不影响读取数据。
至此,MyCAT实现MySQL的读写分离部署测试完毕。
总结:
1. 其实,刚开始配置的是readHost节点,配置如下:
<dataHost name="localhost1" maxCon="1000" minCon="10" balance="1" writeType="0" dbType="mysql" dbDriver="native" switchType="-1" slaveThreshold="100"> <heartbeat>select user()</heartbeat> <!-- can have multi write hosts --> <writeHost host="hostM1" url="localhost:3306" user="root" password="123456"> <!-- can have multi read hosts --> <readHost host="hostS1" url="192.168.244.146:3306" user="root" password="123456" /> </writeHost> </dataHost>
但这种方式有个问题,即master挂了以后,slave也不能提供服务,而这违反了MySQL主从集群的初衷。
2. 如果开启了事务模式,即set autocommit=0,则事务内的读走的是master节点,而不是从节点。