在美国VPS上搭建转发服务，加速服务器上搭建haproxy服务
使用原理: PC->加速服务器frontend->加速服务器backend->美国 VPS
1.安装haproxy：
centos:

yum install haproxy -y

注: 如果yum找不到haproxy的话, 需要下载执行一下:

#CentOS 5 – 32-bit
rpm -Uvh http://mirror.overthewire.com.au/pub/epel/5/i386/epel-release-5-4.noarch.rpm

#CentOS 5 – 64-bit
rpm -Uvh http://mirror.overthewire.com.au/pub/epel/5/x86_64/epel-release-5-4.noarch.rpm

#CentOS 6 – 32-bit
rpm -Uvh http://mirror.overthewire.com.au/pub/epel/6/i386/epel-release-6-8.noarch.rpm

#CentOS 6 – 64-bit
rpm -Uvh http://download.fedoraproject.org/pub/epel/6/x86_64/epel-release-6-8.noarch.rpm

ubuntu

apt-get install haproxy

debian

先添加apt源
vi  /etc/apt/sources.list
然后添加源
deb http://ftp.us.debian.org/debian/ wheezy-backports main
保存后
apt-get update
然后安装haproxy
apt-get install haproxy

2.修改haproxy配置文件，使用脚本启动haproxy

vi /etc/default/haproxy

将该选项修改为1 ENABLED=1

3.修改配置文件

vi /etc/haproxy/haproxy.cfg

单户：

global
ulimit-n 51200
defaults
log global
mode tcp
option dontlognull
contimeout 1000
clitimeout 150000
srvtimeout 150000
frontend ss-in
bind *:8968
default_backend ss-out
backend ss-out
server server1 SS_IP:8968 maxconn 20480  #SS_IP是ss的ip，8968是SS的端口

多户或者均衡

global
ulimit-n  51200
defaults
log global
mode    tcp
option  dontlognull
timeout connect 1000ms
timeout client 150000ms
timeout server 150000ms
frontend 110-in
bind *:110
default_backend 110-out
frontend 109-in
bind *:109
default_backend 109-out
frontend 110-in-jp
bind *:1110
default_backend 110-out-jp
backend 110-out
server server1 spg.xx.xx:110 maxconn 20480
backend 109-out
server server1 spg.xx.xx:109 maxconn 20480
backend 110-out-jp
server server1 jp.xx.xx:110 maxconn 20480

可以配置多端口。
frontend的端口为要在中转服务器上用的端口
backend的端口是源服务器的ip和端口
每个端口的“frontend” 要对应一个“backend”
文件写的时候要记得一一对应。

4.启动服务，检查修改是否生效

service haproxy start

如果前面没设置开机启动，就运行一下：

update-rc.d haproxy defaults

关闭防火墙

sudo ufw disable

借鉴
http://blog.e9china.net/tufan/centos5liyonghaproxyshadowsockszhongzhuan.html
http://www.crazycen.com/internet/2225.html
https://smileawei.com/shadowsocks-haproxy/