HAProxy + Keepalived实现MySQL的高可用负载均衡

dfee

HAProxy+Keepalived实现MySQL的高可用负载均衡

当前环境说明

A:10.10.99.180(MySQLMaster)----|

|________VIP(10.10.99.103)
B:10.10.99.108(MySQLSlave)------|
C:10.10.105.11(MySQLSlave)
D:10.10.105.23(HAProxy+Keepalived)

E:10.10.105.24(HAProxy+Keepalived)
VIP:10.10.105.30

如下操作在10.10.105.23和10.10.105.24上都要进行

1、调整内核参数,添加非本地IP绑定支持

#vi/etc/sysctl.conf
net.ipv4.ip_nonlocal_bind=1
#sysctl–p

2、安装haproxy和keepalived
yum-yinstallhaproxykeepalived

3、配置keepalived.conf文件

#vi/etc/haproxy/haproxy.cfg

!ConfigurationFileforkeepalived

global_defs{

notification_email{

jinyan2049@163.com

}

notification_email_fromkeepalived@chtopnet.com

smtp_server127.0.0.1

smtp_connect_timeout30

router_idLVS_DEVEL

}

vrrp_instanceVI_1{

stateMASTER#10.10.105.24设置为BACKUP

interfaceeth0

virtual_router_id51

realserver10.10.105.23#10.10.105.24改自己的ip

priority90#10.10.105.24设置为80

advert_int1

authentication{

auth_typePASS

auth_pass111111

}

virtual_ipaddress{

10.10.105.30

}

4、配置haproxy

#vi/etc/haproxy/haproxy.cfg

#thisconfigneedshaproxy-1.1.28orhaproxy-1.2.1

global

#log127.0.0.1local0

log127.0.0.1local1notice

maxconn5000

uid99

gid99

daemon

pidfile/var/run/haproxy.pid

defaults

logglobal

modehttp

#optionhttplog

optiondontlognull

retries3

optionredispatch

maxconn2000

contimeout5000

clitimeout50000

srvtimeout50000

listenMYSQL10.10.105.30:3306

modetcp

maxconn2000

balanceroundrobin

servermysql-10.10.99.10810.10.99.108:3306checkinter5000fall1rise2

servermysql-10.10.105.1110.10.105.11:3306checkinter5000fall1rise2

srvtimeout20000

listenstats_auth10.10.105.23:80

#listenstats_auth10.10.105.24:80#backupconfig

statsenable

statsuri/korea

statsauthadmin:12345

statsadminifTRUE

5、分别启动haproxy和keepalived

[iyunv@vm-105-23~]#/etc/init.d/haproxyrestart&&/etc/init.d/keepalivedrestart

Stoppinghaproxy:[OK]

Startinghaproxy:[OK]

Stoppingkeepalived:[OK]

Startingkeepalived:[OK]

6、查看vip是否起来

[iyunv@vm-105-23~]#ipaddr

1:lo:<LOOPBACK,UP,LOWER_UP>mtu16436qdiscnoqueuestateUNKNOWN

link/loopback00:00:00:00:00:00brd00:00:00:00:00:00

inet127.0.0.1/8scopehostlo

inet6::1/128scopehost

valid_lftforeverpreferred_lftforever

2:eth0:<BROADCAST,MULTICAST,UP,LOWER_UP>mtu1500qdiscpfifo_faststateUPqlen1000

link/ether52:54:00:6c:d5:95brdff:ff:ff:ff:ff:ff

inet10.10.105.23/24brd10.10.105.255scopeglobaleth0

inet10.10.105.30/32scopeglobaleth0

inet6fe80::5054:ff:fe6c:d595/64scopelink

valid_lftforeverpreferred_lftforever

7、因为我们知道crontab的颗粒细化度只有1分钟，不能细化到秒，所以我们编写haproxy循环检测脚本，并且放入后台运行

cd/home/ops/scripts

nohupshcheckhapid.sh&

vicheckhapid.sh

#!/bin/bash

whiletrue

do

HA=`ps-ef|grephaproxy|grep-vgrep|wc-l`

if[$HA-eq0];

then

/etc/init.d/haproxystart

echo"">/dev/null

sleep2

if[$HA-eq0];

then

/etc/init.d/keepalivedstop

fi

fi

sleep2

done

8、高可用测试

在10.10.105.23上执行ipaddr|grepeth0

[iyunv@vm-105-23~]#ipaddr|grepeth0

eth0:<BROADCAST,MULTICAST,UP,LOWER_UP>mtu1500qdiscpfifo_faststateUPqlen1000

inet10.10.105.23/24brd10.10.105.255scopeglobaleth0

inet10.10.105.30/32scopeglobaleth0

关闭haproxy再打开一个终端继续执行ipaddr|grepeth0,观测

可以发现当无法迅速启动haproxy进程之后，循环脚本会杀死keepalived进程，实现vip的迁移

10.10.105.30迅速转移到10.10.105.24这台备机上！

9、mysql测试

程序在读写分离之后，所有的select语句可以直接通过访问10.10.105.303306来进行工作，后端的mysql服务器可以实现并发读操作

[iyunv@vm-105-23scripts]#mysqlslap-h10.10.105.30--concurrency=100--iterations=1--create-schema='ultrax'--query='select*frompre_home_share;'--number-of-queries=10--debug-info-ubbs-piz3n3s0ft

Benchmark

Averagenumberofsecondstorunallqueries:1.028seconds

Minimumnumberofsecondstorunallqueries:1.028seconds

Maximumnumberofsecondstorunallqueries:1.028seconds

Numberofclientsrunningqueries:100

Averagenumberofqueriesperclient:0

Usertime0.02,Systemtime0.08

Maximumresidentsetsize6996,Integralresidentsetsize0

Non-physicalpagefaults1584,Physicalpagefaults0,Swaps0

Blocksin0out0,Messagesin0out0,Signals0

Voluntarycontextswitches1280,Involuntarycontextswitches91

10、效果图

可以访问http://10.10.105.23/korea

用户名admin

密码12345

dinggela

花花世界卜必当真、卜管你们吻得多逼真。