如何快速构建高可用集群(Keepalived+Haproxy+Nginx)
转自次元立方网：http://www.it165.net/admin/html/201405/2957.html另一篇：HA-Proxy+Nginx实现web负载均衡（haproxy篇）http://www.it165.net/admin/html/201405/3171.html
组件及实现的功能
Keepalived：实现对Haproxy服务的高可用，并采用双主模型配置;
Haproxy：实现对Nginx的负载均衡和读写分离;
Nginx：实现对HTTP请求的高速处理;
架构设计图
重点概念
vrrp_script中节点权重改变算法
vrrp_script 里的script返回值为0时认为检测成功，其它值都会当成检测失败；
weight 为正时，脚本检测成功时此weight会加到priority上，检测失败时不加；
主失败:
主 priority < 从 priority + weight 时会切换。
主成功：
主 priority + weight > 从 priority + weight 时，主依然为主
weight 为负时，脚本检测成功时此weight不影响priority，检测失败时priority – abs(weight)
主失败:
主 priority – abs(weight) < 从priority 时会切换主从
主成功:
主 priority > 从priority 主依然为主
具体解释详见博文“Keepalived双主模型中vrrp_script中权重改变故障排查”
部署配置
Keepalived部署
配置
001.yum -yinstallkeepalived#两节点都需部署002.# 172.16.25.109003.# vi /etc/keepalived/keepalived.conf004.! Configuration Fileforkeepalived005.global_defs {006. notification_email {007. root@localhost008. }009. notification_email_from admin@lnmmp.com010. smtp_connect_timeout 3011. smtp_server 127.0.0.1012. router_id LVS_DEVEL013.}014.vrrp_script chk_maintaince_down {015. script"[[ -f /etc/keepalived/down]] && exit 1 || exit 0"016. interval 1017. weight 2018.}019.vrrp_script chk_haproxy {020. script"killall -0 haproxy"021. interval 1022. weight 2023.}024.vrrp_instance VI_1 {025. interface eth0026. state MASTER027. priority 100028. virtual_router_id 125029. garp_master_delay 1030. authentication {031. auth_type PASS032. auth_pass 1e3459f77aba4ded033. }034. track_interface {035. eth0036. }037. virtual_ipaddress {038. 172.16.25.10/16 dev eth0 label eth0:0039. }040. track_script {041. chk_haproxy042. }043. notify_master"/etc/keepalived/notify.shmaster 172.16.25.10"044. notify_fault"/etc/keepalived/notify.shfault 172.16.25.10"045.}046.vrrp_instance VI_2 {047. interface eth0048. state BACKUP049. priority 99050. virtual_router_id 126051. garp_master_delay 1052. authentication {053. auth_type PASS054. auth_pass 7615c4b7f518cede055. }056. track_interface {057. eth0058. }059. virtual_ipaddress {060. 172.16.25.11/16 dev eth0 label eth0:1061. }062. track_script {063. chk_haproxy064.chk_maintaince_down065. }066. notify_master"/etc/keepalived/notify.shmaster 172.16.25.11"067. notify_backup"/etc/keepalived/notify.shbackup 172.16.25.11"068. notify_fault"/etc/keepalived/notify.shfault 172.16.25.11"069.}070.# 172.16.25.110071.# vi /etc/keepalived/keepalived.conf072.! Configuration Fileforkeepalived073.global_defs {074. notification_email {075. root@localhost076. }077. notification_email_from admin@lnmmp.com078. smtp_connect_timeout 3079. smtp_server 127.0.0.1080. router_id LVS_DEVEL081.}082.vrrp_script chk_maintaince_down {083. script"[[ -f /etc/keepalived/down]] && exit 1 || exit 0"084. interval 1085. weight 2086.}087.vrrp_script chk_haproxy {088. script"killall -0 haproxy"089. interval 1090. weight 2091.}092.vrrp_instance VI_1 {093. interface eth0094. state BACKUP095. priority 99096. virtual_router_id 125097. garp_master_delay 1098. authentication {099. auth_type PASS100. auth_pass 1e3459f77aba4ded101. }102. track_interface {103. eth0104. }105. virtual_ipaddress {106. 172.16.25.10/16 dev eth0 label eth0:0107. }108. track_script {109. chk_haproxy110.chk_maintaince_down111. }112. notify_master"/etc/keepalived/notify.shmaster 172.16.25.10"113. notify_backup"/etc/keepalived/notify.shbackup 172.16.25.10"114. notify_fault"/etc/keepalived/notify.shfault 172.16.25.10"115.}116.vrrp_instance VI_2 {117. interface eth0118. state MASTER119. priority 100120. virtual_router_id 126121. garp_master_delay 1122. authentication {123. auth_type PASS124. auth_pass 7615c4b7f518cede125. }126. track_interface {127. eth0128. }129. virtual_ipaddress {130. 172.16.25.11/16 dev eth0 label eth0:1131. }132. track_script {133. chk_haproxy134. }135. notify_master"/etc/keepalived/notify.shmaster 172.16.25.11"136. notify_backup"/etc/keepalived/notify.shbackup 172.16.25.11"137. notify_fault"/etc/keepalived/notify.shfault 172.16.25.11"138.}139.# vi /etc/keepalived/notify.sh140.#!/bin/bash141.# Author: Jason.Yu <admin@lnmmp.com>142.# description: An example of notify script143.#144.contact='root@localhost'145.notify() {146. mailsubject="`hostname` tobe $1: $2 floating"147. mailbody="`date '+%F %H:%M:%S'`:vrrp transition, `hostname` changed to be $1"148. echo$mailbody | mail-s "$mailsubject"$contact149.}150.case"$1"in151. master)152. notify master $2153. /etc/rc.d/init.d/haproxy restart154. exit0155. ;;156. backup)157. notify backup $2# 在节点切换成backup状态时，无需刻意停止haproxy服务，防止chk_maintaince和chk_haproxy多次对haproxy服务操作；158. exit0159. ;;160. fault)161. notify fault $2# 同上162. exit0163. ;;164. *)165. echo'Usage: `basename$0` {master|backup|fault}' 166. exit1167. ;;168.esac启动服务
1.service keepalived start# 在两个节点上都需要启动keepalived双主模型启动
Haproxy部署
安装配置
01.yum -yinstallhaproxy#两节点都需部署02.vi/etc/haproxy/haproxy.cfg#两节点配置一致03.global04. log 127.0.0.1 local205. chroot /var/lib/haproxy06. pidfile /var/run/haproxy.pid07. maxconn 400008. user haproxy09. group haproxy10. daemon# 以后台程序运行；11.defaults12. mode http#选择HTTP模式，即可进行7层过滤； 13. log global14. option httplog#可以得到更加丰富的日志输出； 15. option dontlognull16. option http-server-close#server端可关闭HTTP连接的功能； 17. option forwardfor except 127.0.0.0/8#传递client端的IP地址给server端，并写入“X-Forward_for”首部中；18. option redispatch19. retries 320. timeout http-request 10s21. timeout queue 1m22. timeout connect 10s23. timeout client 1m24. timeout server 1m25. timeout http-keep-alive 10s26. timeout check 10s27. maxconn 3000028.listen stats29. modehttp30. bind0.0.0.0:1080# 统计页面绑定1080端口；31. statsenable#开启统计页面功能； 32. stats hide-version# 隐藏Haproxy版本号；33. stats uri /haproxyadmin?stats#自定义统计页面的访问uri； 34. stats realm Haproxy\ Statistics#统计页面密码验证时的提示信息； 35. stats auth admin:admin#为统计页面开启登录验证功能； 36. stats adminifTRUE#若登录用户验证通过，则赋予管理功能；37.frontend http-in38. bind*:8039. mode http40. log global41. option httpclose42. option logasap43. option dontlognull44. capture requestheader Host len 2045. capture requestheader Referer len 6046. acl url_static path_beg -i /static /images /javascript /stylesheets47. acl url_static path_end -i .jpg .jpeg .gif .png .css .js .html48. use_backend static_serversifurl_static#符合ACL规则的，请求转入后端静态服务器49. default_backend dynamic_servers#默认请求转入后端动态服务器 50.backend static_servers51. balance roundrobin52. server imgsrv1 192.168.0.25:80 check maxconn 6000#静态服务器，可配置多台，还可设置权重weight；53.backend dynamic_servers54. balancesource#对于动态请求利用source调度算法，可一定程度上实现session保持；但最好利用cookie绑定的方式实现session保持55. server websrv1 192.168.0.35:80 check maxconn 1000#动态服务器，可配置多台，还可设置权重weight；启动服务
1.service haproxy start# 两节点都需要启动Nginx部署
见博客“如何测试Nginx的高性能”http://www.it165.net/admin/html/201405/2928.html
访问验证
Haproxy统计页面测试
QTNTdG14REFCVy14MHUxTUYwMjA5LmpwZw==" border="0" height="400" hspace="0" src="http://www.it165.net/uploadfile/files/2014/0504/20140504094357372.jpg"title="统计页面展示.png" vspace="0" width="600" />
动静分离测试
高可用测试
页:
[1]