一、MySQL 5.6介绍


MySQL 5.6引入的GTID(Global Transaction IDs)使得其复制功能的配置、监控及管理变得更加易于实现,且更加健壮。

复制功能的改进:

1、支持多线程复制,指的是针对每个database开启相应的独立线程。

2、支持启用GTID,就是每个mysql服务器都一个全局的事务ID,GTID,这个是由每个服务器的UUID(唯一的)和server_id组成的,mysql会保证事务与GTID之间1:1的映射,保证了GTID的全局唯一性。


二、Amoeba介绍

Amoeba是一个SQL请求的路由器,可以为mysql提供负载均衡、读写分离、高可用性提供机制,用户需要结合使用MySQL的Replication等机制来实现副本同步等功能,它主要在应用层访问MySQL的时候充当query 路由功能,专注分布式数据库 proxy 开发,座落与Client、DB Server(s)之间,对客户端透明,它是一个虚拟的mysql,对外提供mysql协议,客户端连接amoeba就象连接mysql一样。


Amoeba主要优势:

1、数据切分后复杂数据源整合;

2、提供数据切分规则并降低数据切分规则给数据库带来的影响;

3、降低数据库与客户端连接;

4、读写分离路由。

Amoeba不足:

1、不支持事务;

2、不支持存储过程;

3、不适合从amoeba导数据的场景或者对大数据量查询的query并不合适;

4、不支持分库分表,amoeba只支持分数据库实例,每个被切分的节点需要保持库表结构一致。


三、实验环境


192.168.30.115 OS:CentOS 6.4 x86_64  amoeba.luojianlong.com

192.168.30.116 OS:CentOS 6.4 x86_64  master.luojianlong.com

192.168.30.117 OS:CentOS 6.4 x86_64  slave.luojianlong.com

MySQL version:mysql-5.6.13-linux-glibc2.5-x86_64

Amoeba version:amoeba-mysql-binary-2.1.0-RC5


拓扑图如下:

MySQL 5.6.13 GTID复制,以及基于Amoeba实现读写分离_第1张图片




首先,分别在master.luojianlong.com和slave.luojianlong.com上安装mysql 5.6.13,这里使用通用的二进制包来安装

[root@master ~]# useradd -r -u 27 mysql
[root@master ~]# tar zxvf mysql-5.6.13-linux-glibc2.5-x86_64.tar.gz -C /usr/local/
[root@master ~]# cd /usr/local/
[root@master local]# ln -s mysql-5.6.13-linux-glibc2.5-x86_64 mysql
[root@master local]# ln -s mysql-5.6.13-linux-glibc2.5-x86_64 mysql
[root@master local]# cd mysql
[root@master mysql]# mkdir /mydata/data -p
[root@master mysql]# chown -R root.mysql ./*
[root@master mysql]# chown -R mysql.mysql /mydata/data/
[root@master mysql]# ./scripts/mysql_install_db --user=mysql --datadir=/mydata/data
[root@master mysql]# cp support-files/mysql.server /etc/rc.d/init.d/mysqld
[root@master mysql]# chkconfig --add mysqld
[root@master mysql]# chkconfig mysqld on
[root@master mysql]# chmod +x /etc/rc.d/init.d/mysqld
[root@master ~]# vi /etc/profile.d/mysql.sh
export PATH=/usr/local/mysql/bin:$PATH
[root@master ~]# . /etc/profile.d/mysql.sh
# 由于mysql 5.6版本默认的配置文件配置参数很少,需要自己定制,为了方便,这里使用mysql 5.5的配置文件
[root@master mysql]# cp my-large.cnf /etc/my.cnf
[root@master mysql]# vi /etc/my.cnf
# 在[mysqld]中添加如下
datadir = /mydata/data
innodb_file_per_table = 1
binlog-format=ROW
log-bin=master-bin
log-slave-updates=true
gtid-mode=on
enforce-gtid-consistency=true
master-info-repository=TABLE
relay-log-info-repository=TABLE
sync-master-info=1
slave-parallel-workers=2
binlog-checksum=CRC32
master-verify-checksum=1
slave-sql-verify-checksum=1
binlog-rows-query-log_events=1
server-id=1
report-port=3306
port=3306
report-host=192.168.30.116


在slave.luojianlong.com中也使用相同的方式安装mysql 5.6

# 安装完成后,修改配置文件
[root@slave ~]# vi /etc/my.cnf
# 添加如下
datadir = /mydata/data
innodb_file_per_table = 1
binlog-format=ROW
log-slave-updates=true
gtid-mode=on
enforce-gtid-consistency=true
master-info-repository=TABLE
relay-log-info-repository=TABLE
sync-master-info=1
slave-parallel-workers=2
binlog-checksum=CRC32
master-verify-checksum=1
slave-sql-verify-checksum=1
binlog-rows-query-log_events=1
server-id=11
report-port=3306
port=3306
log-bin=mysql-bin.log
datadir=/mydata/data
report-host=192.168.30.117


binlog-format:二进制日志的格式,有row、statement和mixed几种类型;

需要注意的是:当设置隔离级别为READ-COMMITED必须设置二进制日志格式为ROW,现在MySQL官方认为STATEMENT这个已经不再适合继续使用;但mixed类型在默认的事务隔离级别下,可能会导致主从数据不一致;

log-slave-updates、gtid-mode、enforce-gtid-consistency、report-port和report-host:用于启动GTID及满足附属的其它需求;

master-info-repository和relay-log-info-repository:启用此两项,可用于实现在崩溃时保证二进制及从服务器安全的功能;

sync-master-info:启用之可确保无信息丢失;

slave-paralles-workers:设定从服务器的SQL线程数;0表示关闭多线程复制功能;

binlog-checksum、master-verify-checksum和slave-sql-verify-checksum:启用复制有关的所有校验功能;

binlog-rows-query-log-events:启用之可用于在二进制日志记录事件相关的信息,可降低故障排除的复杂度;

log-bin:启用二进制日志,这是保证复制功能的基本前提;

server-id:同一个复制拓扑中的所有服务器的id号必须惟一;



分别启动俩台服务器的mysql

[root@master ~]# service mysqld start
Starting MySQL............ SUCCESS!
[root@slave ~]# service mysqld start
Starting MySQL... SUCCESS!


在master服务器上创建复制用户

mysql> grant replication slave on *.* to 'repluser'@'192.168.30.%' identified by '123456';
Query OK, 0 rows affected (0.06 sec)
mysql> flush privileges;
Query OK, 0 rows affected (0.00 sec)


启动从节点的复制线程

mysql> change master to master_host='192.168.30.116', master_user='repluser', master_password='123456', master_auto_position=1;
Query OK, 0 rows affected, 2 warnings (0.45 sec)
mysql> start slave;
Query OK, 0 rows affected, 1 warning (0.12 sec)


在slave上查看线程启动情况

mysql> show slave status\G;
*************************** 1. row ***************************
               Slave_IO_State: Waiting for master to send event
                  Master_Host: 192.168.30.116
                  Master_User: repluser
                  Master_Port: 3306
                Connect_Retry: 60
              Master_Log_File: master-bin.000002
          Read_Master_Log_Pos: 539
               Relay_Log_File: slave-relay-bin.000002
                Relay_Log_Pos: 751
        Relay_Master_Log_File: master-bin.000002
             Slave_IO_Running: Yes
            Slave_SQL_Running: Yes
              Replicate_Do_DB:
          Replicate_Ignore_DB:
           Replicate_Do_Table:
       Replicate_Ignore_Table:
      Replicate_Wild_Do_Table:
  Replicate_Wild_Ignore_Table:
                   Last_Errno: 0
                   Last_Error:
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 539
              Relay_Log_Space: 955
              Until_Condition: None
               Until_Log_File:
                Until_Log_Pos: 0
           Master_SSL_Allowed: No
           Master_SSL_CA_File:
           Master_SSL_CA_Path:
              Master_SSL_Cert:
            Master_SSL_Cipher:
               Master_SSL_Key:
        Seconds_Behind_Master: 0
Master_SSL_Verify_Server_Cert: No
                Last_IO_Errno: 0
 Last_IO_Error:
               Last_SQL_Errno: 0
               Last_SQL_Error:
  Replicate_Ignore_Server_Ids:
             Master_Server_Id: 1
                  Master_UUID: b05c9b5c-bfac-11e3-a4bc-000c29f3fcba
             Master_Info_File: mysql.slave_master_info
                    SQL_Delay: 0
          SQL_Remaining_Delay: NULL
      Slave_SQL_Running_State: Slave has read all relay log; waiting for the slave I/O thread to update it
           Master_Retry_Count: 86400
                  Master_Bind:
      Last_IO_Error_Timestamp:
     Last_SQL_Error_Timestamp:
               Master_SSL_Crl:
           Master_SSL_Crlpath:
           Retrieved_Gtid_Set: b05c9b5c-bfac-11e3-a4bc-000c29f3fcba:1-2
            Executed_Gtid_Set: b05c9b5c-bfac-11e3-a4bc-000c29f3fcba:1-2
                Auto_Position: 1
1 row in set (0.00 sec)


Connect_Retry:slave服务器连接master服务器的重试次数;

Exec_Master_Log_Pos:当前slave服务器最后一次所执行二进制日志的位置;

Relay_Log_Space:relaylog的空间;

Seconds_Behind_Master:slave滞后master的时长(通过当前服务器的时间戳与二进制日志中的时间戳对比得到的);

Slave_IO_Running:slave服务器I/O线程的运行情况;

Slave_SQL_Running:slave服务器SQL线程的运行情况;


发现gtid复制已经启动成功,并可以查看主从服务器的uuid(对于每一个mysql服务器都是唯一的)

mysql> show global variables like '%uuid';
+---------------+--------------------------------------+
| Variable_name | Value                                |
+---------------+--------------------------------------+
| server_uuid   | b05c9b5c-bfac-11e3-a4bc-000c29f3fcba |
+---------------+--------------------------------------+
1 row in set (0.00 sec)
mysql> show global variables like '%uuid';
+---------------+--------------------------------------+
| Variable_name | Value                                |
+---------------+--------------------------------------+
| server_uuid   | ba83181a-bfac-11e3-a4bc-000c295b50f9 |
+---------------+--------------------------------------+
1 row in set (0.00 sec)

在slave 可以看到启动的复制线程数

mysql> show processlist;
+-----+-------------+-----------+------+---------+-------+-----------------------------------------------------------------------------+------------------+
| Id  | User        | Host      | db   | Command | Time  | State                                                                       | Info             |
+-----+-------------+-----------+------+---------+-------+-----------------------------------------------------------------------------+------------------+
|   3 | system user |           | NULL | Connect | 67092 | Waiting for master to send event                                            | NULL             |
|   4 | system user |           | NULL | Connect | 50955 | Slave has read all relay log; waiting for the slave I/O thread to update it | NULL             |
|   5 | system user |           | NULL | Connect | 65267 | Waiting for an event from Coordinator                                       | NULL             |
|   6 | system user |           | NULL | Connect | 50961 | Waiting for an event from Coordinator                                       | NULL             |
| 151 | root        | localhost | NULL | Query   |     0 | init                                                                        | show processlist |
+-----+-------------+-----------+------+---------+-------+-----------------------------------------------------------------------------+------------------+
5 rows in set (0.00 sec)


在master创建数据库,看是否同步到slave

mysql> create database student;
Query OK, 1 row affected (0.01 sec)
mysql> show databases;
+--------------------+
| Database           |
+--------------------+
| information_schema |
| abc                |
| mysql              |
| performance_schema |
| student            |
| test               |
| testdb             |
+--------------------+
7 rows in set (0.00 sec)

发现已经同步到slave


主从服务器数据不一致的常见原因:

1、在slave服务器上直接执行写操作;

2、基于SQL语句的复制,建议使用基于数据本身的复制;

3、relaylog日志出现故障;

4、数据库架构复制有问题;

5、在一个数据库里面混合使用myisam和innodb引擎;

6、mysql的主从版本不一致(从服务器的版本必须高于主服务器版本);

7、使用myisam存储引擎时,执行的语句被意外终止;


修复主从服务器数据不一致的方案:

1、使用pt-table-checksum工具,能够监控主从服务器数据表中的数据是否一致;

2、使用pt-table-sync工具,能够实现同步主从服务器上表的数据;


防止主从复制崩溃后带来的损失:

1、设置master服务器的sync_binlog=1,意思是只要有数据更新就立即同步到磁盘上面,同时要求磁盘本身不具有缓存能力;

2、使用innodb存储引擎时,当slave服务器意外关闭后,重新启动会重新读取master_info文件,这个时候master_info数据没刷新到磁盘上面,这样会导致数据复制出现错误,这个时候可以在slave上启用innodb_overwrite_relay_log_info;可以覆盖relay_log_info里面的数据,避免造成数据丢失;

3、或者也可以使用工具pt-slave-restart,从节点意外关闭时,用于安全启动复制进程;

4、当master服务器二进制日志出现错误时,可以在slave服务器上执行set global sql_slave_skip_counter=1;start slave; 跳过错误,或者删除二进制日志重新生成,然后重新建立复制;



下面开始配置amoeba,首先安装jdk环境

[root@amoeba ~]# tar zxvf jdk-7u25-linux-x64.gz -C /usr/local/
[root@amoeba ~]# cd /usr/local/
[root@amoeba local]# ln -s jdk1.7.0_25 jdk
[root@amoeba local]# vi /etc/profile.d/jdk.sh
export JAVA_HOME=/usr/local/jdk
export CLASSPATH=$CLASSPATH:$JAVA_HOME/lib:$JAVA_HOME/jre/lib
export PATH=/usr/local/jdk/bin:$PATH
[root@amoeba local]# . /etc/profile.d/jdk.sh
[root@amoeba local]# java -version
java version "1.7.0_25"
Java(TM) SE Runtime Environment (build 1.7.0_25-b15)
Java HotSpot(TM) 64-Bit Server VM (build 23.25-b01, mixed mode)


下载安装amoeba

[root@amoeba ~]# mkdir /usr/local/amoeba
[root@amoeba ~]# tar zxvf amoeba-mysql-binary-2.1.0-RC5.tar.gz -C /usr/local/amoeba/
[root@amoeba ~]# vi /etc/profile.d/amoeba.sh
export PATH=/usr/local/amoeba/bin:$PATH
[root@amoeba ~]# . /etc/profile.d/amoeba.sh
[root@amoeba ~]# cp /usr/local/amoeba/conf/amoeba.xml /usr/local/amoeba/conf/amoeba.xml.bak
[root@amoeba ~]# cp /usr/local/amoeba/conf/dbServers.xml /usr/local/amoeba/conf/dbServers.xml.bak

调整amoeba配置文件,设置读写分离

[root@amoeba ~]# vi /usr/local/amoeba/conf/amoeba.xml
# 在子标签中修改如下几项
3306
0.0.0.0
mypass
# 在子标签中启用修改下面几项
server1
server2
[root@amoeba ~]# vi /usr/local/amoeba/conf/dbServers.xml
# 启动并修改 子标签中的
mypass
# 修改 子标签中的
192.168.30.116
# 新添加一个dbServer如下

                
                        
                        192.168.30.117
                
        


在master和slave中分别创建用于amoeba连接俩台mysql的用户

# 在master上面创建
mysql> grant all on *.* to 'root'@'192.168.30.%' identified by 'mypass';
Query OK, 0 rows affected (0.00 sec)
mysql> flush privileges;
Query OK, 0 rows affected (0.00 sec)
#发现已经同步到slave
mysql> show grants for 'root'@'192.168.30.%';
+-------------------------------------------------------------------------------------------------------------------------+
| Grants for [email protected].%                                                                                            |
+-------------------------------------------------------------------------------------------------------------------------+
| GRANT ALL PRIVILEGES ON *.* TO 'root'@'192.168.30.%' IDENTIFIED BY PASSWORD '*6C8989366EAF75BB670AD8EA7A7FC1176A95CEF4' |
+-------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)


启动amoeba

[root@amoeba ~]# vi /usr/local/amoeba/bin/amoeba
DEFAULT_OPTS="-server -Xms256m -Xmx256m -Xss128k"
# 修改为
DEFAULT_OPTS="-server -Xms256m -Xmx256m -Xss256k"
[root@amoeba ~]# amoeba start
log4j:WARN log4j config load completed from file:/usr/local/amoeba/conf/log4j.xml
2014-04-09 15:09:15,121 INFO  context.MysqlRuntimeContext - Amoeba for Mysql current versoin=5.1.45-mysql-amoeba-proxy-2.1.0-RC5
log4j:WARN ip access config load completed from file:/usr/local/amoeba/conf/access_list.conf
2014-04-09 15:09:15,482 INFO  net.ServerableConnectionManager - Amoeba for Mysql listening on /0.0.0.0:3306.
2014-04-09 15:09:15,489 INFO  net.ServerableConnectionManager - Amoeba Monitor Server listening on /127.0.0.1:53201.
[root@amoeba ~]# netstat -anptl | grep :3306
tcp        0      0 :::3306                     :::*                        LISTEN      19208/java
tcp        0      0 ::ffff:192.168.30.115:33408 ::ffff:192.168.30.116:3306  ESTABLISHED 19208/java
tcp        0      0 ::ffff:192.168.30.115:59880 ::ffff:192.168.30.117:3306  ESTABLISHED 19208/java
[root@amoeba ~]# mysql -u root -pmypass -h 192.168.30.115 -P 3306
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 1783342624
Server version: 5.1.45-mysql-amoeba-proxy-2.1.0-RC5 MySQL Community Server (GPL)
Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
Oracle is a registered trademark of Oracle Corporation and/or its
affiliates. Other names may be trademarks of their respective
owners.
Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.
mysql> show databases;
+--------------------+
| Database           |
+--------------------+
| information_schema |
| mysql              |
| performance_schema |
| test               |
+--------------------+
4 rows in set (0.03 sec)


启动成功,并访问正常


下面使用tcpdump命令抓包测试读写分离效果

[root@master ~]# tcpdump -i eth0 -s0 -nn -A tcp dst port 3306 and dst host 192.168.30.116
[root@slave ~]# tcpdump -i eth0 -s0 -nn -A tcp dst port 3306 and dst host 192.168.30.117
# amoeba服务器上登录mysql,并创建数据库
mysql> create database abc;
Query OK, 1 row affected (0.03 sec)
[root@master ~]# tcpdump -i eth0 -s0 -nn -A tcp dst port 3306 and dst host 192.168.30.116
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 65535 bytes
15:16:42.492828 IP 192.168.30.115.33408 > 192.168.30.116.3306: Flags [P.], seq 1149949367:1149949391, ack 113323290, win 123, options [nop,nop,TS val 8373727 ecr 10837334], length 24
E..L&"@[email protected]....{4].....
......]V.....create database abc
15:16:42.501837 IP 192.168.30.115.33408 > 192.168.30.116.3306: Flags [.], ack 12, win 123, options [nop,nop,TS val 8373736 ecr 11221763], length 0
E..4&#@[email protected]%...{.I.....
......;.
15:16:42.502356 IP 192.168.30.117.38626 > 192.168.30.116.3306: Flags [.], ack 3194819787, win 170, options [nop,nop,TS val 9637577 ecr 11221763], length 0
E..4.0@[email protected].............
......;.
# 发现在master上面抓到了创建数据库的包
[root@slave ~]# tcpdump -i eth0 -s0 -nn -A tcp dst port 3306 and dst host 192.168.30.117
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 65535 bytes
# slave上面没有抓到任何包
# 下面在amoeba上面执行创建表,插入,查询语句
mysql> use abc
Database changed
mysql> create table test(id int);
Query OK, 0 rows affected (0.41 sec)
mysql> insert into test values(1);
Query OK, 1 row affected (0.04 sec)
mysql> select * from test;
+------+
| id   |
+------+
|    1 |
+------+
1 row in set (0.01 sec)
[root@master ~]# tcpdump -i eth0 -s0 -nn -A tcp dst port 3306 and dst host 192.168.30.116
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 65535 bytes
15:20:42.054548 IP 192.168.30.115.33419 > 192.168.30.116.3306: Flags [P.], seq 976065559:976065590, ack 2591052457, win 123, options [nop,nop,TS val 8613289 ecr 11418139], length 31
E..S."@[email protected]....:-...pR....{.H.....
..m...:......insert into test values(1)
15:20:42.085305 IP 192.168.30.117.38626 > 192.168.30.116.3306: Flags [.], ack 3194820258, win 189, options [nop,nop,TS val 9877161 ecr 11461346], length 0
E..4.2@[email protected].......*.....
........
15:20:42.086686 IP 192.168.30.115.33419 > 192.168.30.116.3306: Flags [.], ack 12, win 123, options [nop,nop,TS val 8613320 ecr 11461347], length 0
E..4.#@[email protected]....:-.6.pR....{./.....
..m.....
[root@slave ~]# tcpdump -i eth0 -s0 -nn -A tcp dst port 3306 and dst host 192.168.30.117
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on eth0, link-type EN10MB (Ethernet), capture size 65535 bytes
15:20:52.076327 IP 192.168.30.115.59899 > 192.168.30.117.3306: Flags [P.], seq 201259223:201259232, ack 2661648116, win 115, options [nop,nop,TS val 8623261 ecr 9819401], length 9
E..=..@[email protected].&.....
....... .....abc.
15:20:52.076821 IP 192.168.30.115.59899 > 192.168.30.117.3306: Flags [.], ack 12, win 115, options [nop,nop,TS val 8623261 ecr 9887103], length 0
E..4..@[email protected]#q.....
........
15:20:52.077609 IP 192.168.30.115.59899 > 192.168.30.117.3306: Flags [P.], seq 9:32, ack 12, win 115, options [nop,nop,TS val 8623262 ecr 9887103], length 23
E..K..@[email protected].......
.............select * from test
15:20:52.118046 IP 192.168.30.115.59899 > 192.168.30.117.3306: Flags [.], ack 82, win 115, options [nop,nop,TS val 8623303 ecr 9887105], length 0
E..4..@[email protected]"......
观察抓包情况,发现所有的更新操作都在master,查询操作都在slave 上面


登录主从数据库查看数据是否一致

# master
mysql> use abc
Database changed
mysql> select * from test;
+------+
| id   |
+------+
|    1 |
+------+
1 row in set (0.00 sec)
# slave
mysql> use abc
Database changed
mysql> select * from test;
+------+
| id   |
+------+
|    1 |
+------+
1 row in set (0.00 sec)
# 发现数据一致



四、Amoeba简单的优化配置:

1、 修改amoeba启动参数,增加JAVA虚拟机的内存,修改amoeba执行程序,查找DEFAULT_OPTS,修改为以下内容:

DEFAULT_OPTS="-server -Xms1024m -Xmx1024m -Xss256k"

2、 amoeba.xml中关于多线程和网络的优化

[root@amoeba ~]# vim /usr/local/amoeba/conf/amoeba.xml
# 修改如下几项
500
500
500
256
128


3、 amoeba.xml中还有一个处理客户端连接和网络信息使用CPU核数的配置,不过默认配置是主机拥有的CPU核的数量,一般不用修改,如果你的amoeba负载较高,并且机器上还有其他服务,你就可以降低这个配置,避免系统或者其他服务没有资源,或者出现资源争抢,从而导致整理的处理能力下降。

查找amoeba.xml文件中connectionManagerList关键字,修改

[root@amoeba ~]# vi /usr/local/amoeba/conf/amoeba.xml

# 取消注释,并设置processors数量。


到此,MySQL 5.6基于GTID的复制,以及Amoeba读写分离的实现配置完成。