Nginx+Keepalived简单构建高可用集群

以前一直用heartbeat或corosync+pacemaker构建高可用集群，现在发现keepalived实现起来更简单。
keepalived的master向backup发送广播，当backup一段时间收不到对方传来的VRRP广播时，backup会通过竞选一个master,master就会重新持有资源。具体的理论知识参见http://bbs.ywlm.net/thread-790-1-1.html

实验目标:2台Nginx+Keepalived 2台Lamp构建高可用Web集群

规划：

ng1.laoguang.me 192.168.1.22 ng1
ng2.laoguang.me 192.168.1.23 ng2
lamp1.laoguang.me   192.168.1.24 lamp1
lamp2.laoguang.me   192.168.1.25 lamp2

拓扑：

一.基本环境准备
ng1,ng2上安装nginx
lamp1,lamp2上构建LAMP或只安装httpd，我只安装了Httpd，这里不给大家演示了，有需要请看我的其它博文，更改lamp1,lamp2的index.html的内容分别为lamp1和lamp2,以容易区分，实际集群中内容应该是一致的，由共享存储提供。

二.ng1,ng2上安装配置keepalived
下载地址:http://www.keepalived.org/download.html
2.1 安装keepalived

tar xvf keepalived-1.2.7.tar.gz
cd keepalived-1.2.7
./configure --prefix=/usr/local/keepalived
##可能会提示安装popt-devel包，yum即可
make && make install

2.2 整理配置文件与脚本

mkdir /etc/keepalived
##keepalived默认配置文件从/etc/keepalived下读取
cp /usr/local/keepalived/sbin/keepalived /usr/sbin/
##就一个二进制文件，直接拷贝过去即可，多的话就更改PATH吧
cp /usr/local/keepalived/etc/sysconfig/keepalived /etc/sysconfig/
##脚本的额外配置文件读取位置
cp /usr/local/keepalived/etc/rc.d/init.d/keepalived /etc/init.d/
##启动脚本你懂得
cp /usr/local/keepalived/etc/keepalived/keepalived.conf /etc/keepalived/
##我们关键的keepalived配置文件

2.3 修改ng1的/etc/keepalived/keepalived.conf

! Configuration File for keepalived
global_defs {
   notification_email {
     [email protected]         ##出故障发送邮件给谁
   }
   notification_email_from keepalived@localhost ##故障用哪个邮箱发送邮件
   smtp_server 127.0.0.1   ##SMTP_Server IP
   smtp_connect_timeout 30 ##超时时间
   router_id LVS_DEVEL     ##服务器标识
}
vrrp_instance VI_1 {
    state BACKUP
##状态，都为BACKUP，它们会推选Master，如果你写MASTER，它就会是Master，
    ##当Master故障时Backup会成为Master,当原来的Master恢复后，原来的Master会成为Master
    interface eth0        ##发送VRRP的接口，仔细看你的是不是eth0
    virtual_router_id 51  ##虚拟路由标识，同一个组应该用一个，即Master与Backup同一个
    priority 100   ##重要的优先级哦
    nopreempt      ##不抢占，一个故障时，重启后恢复后不抢占意资源
    advert_int 1   ##同步间隔时长
    authentication {              ##认证
        auth_type PASS            ##认证方式
        auth_pass www.laoguang.me ##密钥
    }
    virtual_ipaddress {
        192.168.1.18/24 dev eth0              ##VIP
    }
}
##后面的删除吧，LVS上才有用

拷贝到ng2上一份，只修改priority 90 即可

scp /etc/keepalived/keepalived.conf 192.168.1.23:/etc/keepalived/
##Ng2上
vi /etc/keepalived/keepalived.conf  priority 90   ##其它一致

2.4 ng1,ng2上启动keepalived

service keepalived start

查看日志

tail /var/log/messages
Nov 27 08:07:54 localhost Keepalived_vrrp[41871]: VRRP_Instance(VI_1) Entering BACKUP STATE
Nov 27 08:07:54 localhost Keepalived_vrrp[41871]: VRRP sockpool: [ifindex(2), proto(112), fd(11,12)]
Nov 27 08:07:54 localhost Keepalived_healthcheckers[41870]: Using LinkWatch kernel netlink reflector...
Nov 27 08:07:54 localhost Keepalived_vrrp[41871]: VRRP_Instance(VI_1) forcing a new MASTER election
Nov 27 08:07:55 localhost Keepalived_vrrp[41871]: VRRP_Instance(VI_1) Transition to MASTER STATE
Nov 27 08:07:56 localhost Keepalived_vrrp[41871]: VRRP_Instance(VI_1) Entering MASTER STATE
Nov 27 08:07:56 localhost Keepalived_vrrp[41871]: VRRP_Instance(VI_1) setting protocol VIPs.
Nov 27 08:07:56 localhost Keepalived_healthcheckers[41870]: Netlink reflector reports IP 192.168.1.18 added
Nov 27 08:07:56 localhost Keepalived_vrrp[41871]: VRRP_Instance(VI_1) Sending gratuitous ARPs on eth0 for 192.168.1.18
Nov 27 08:08:01 localhost Keepalived_vrrp[41871]: VRRP_Instance(VI_1) Sending gratuitous ARPs on eth0 for 192.168.1.18

查看vip绑定到哪台机器上了

ip addr     ##ng1上
....省略
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:e8:90:0b brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.22/24 brd 192.168.1.255 scope global eth0
    inet 192.168.1.18/32 scope global eth0
    inet6 fe80::20c:29ff:fee8:900b/64 scope link
       valid_lft forever preferred_lft forever

由此可知vip绑定到ng1上了
三，Keepalived测试

3.1 关闭ng1上的keepalived或者直接关闭ng1 查看vip转移情况

service keepalived stop
ip addr
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:e8:90:0b brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.22/24 brd 192.168.1.255 scope global eth0
    inet6 fe80::20c:29ff:fee8:900b/64 scope link
       valid_lft forever preferred_lft forever

3.2 查看ng2上是否绑定了vip

ip addr
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
    link/ether 00:0c:29:dd:00:77 brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.23/24 brd 192.168.1.255 scope global eth0
    inet 192.168.1.18/32 scope global eth0
    inet6 fe80::20c:29ff:fedd:77/64 scope link
       valid_lft forever preferred_lft forever

由此可知ip转移正常，keepalived设置成功

四.配置Nginx做反向代理

4.1 修改nginx配置文件

vi /etc/nginx/nginx.conf
user  nginx nginx;   ##运行nginx的用户和组
worker_processes  2; ##启动进程数
error_log /var/log/nginx/error.log  notice;  ##错误日志记录
pid        /tmp/nginx.pid;                   ##pid存放位置
worker_rlimit_nofile 65535;                  ##线程最大打开文件数，须配合ulimit -SHn使用
events {
    use epoll;                 ##工作模型
    worker_connections  65536; ##单进程最大连接数
}
http {                         ##http模块
    include       mime.types;  ##包含进来
    default_type  application/octet-stream; ##默认类型
    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" '
                      '$status $body_bytes_sent "$http_referer" '
                      '"$http_user_agent" "$http_x_forwarded_for"';
##日志格式
    access_log  /var/logs/nginx/http.access.log  main; ##访问日志
    client_max_body_size 20m;  ##最大请求文件大小
    client_header_buffer_size 16k; ##来自客户端请求header_buffer大小
    large_client_header_buffers 4 16k; ##较大请求缓冲个数与大小
    sendfile       on;                 ##内核空间直接发送到tcp队列
    tcp_nopush     on;
    tcp_nodelay    on;
    keepalive_timeout  65;     ##长连接时长
    gzip  on;                  ##启用压缩
    gzip_min_length 1k;        ##最小压缩大小
    gzip_buffers 4 16k;        ##压缩缓冲
    gzip_http_version 1.1;     ##支持协议
    gzip_comp_level 2;         ##压缩等级
    gzip_types text/plain application/x-javascript text/css application/xml;      ##压缩类型
    gzip_vary on;              ##前端缓存服务器可以缓存压缩过的页面
    upstream laoguang.me {     ##用upstream模块定义集群与RS
        server 192.168.1.24:80 max_fails=3fail_timeout=10s;   ##RS的地址，最大错误数与超时时间，超过了自动剔除
        server 192.168.1.25:80 max_fails=3fail_timeout=10s;
}
 server {
        listen       80;           ##监听端口
        server_name  192.168.1.18; ##servername
        root   html;               ##根目录
        index  index.html index.htm; ##你懂得
        #charset koi8-r;
        access_log  logs/192.168.1.18.access.log  main;
  ##这个server的访问日志
        location / {
                proxy_pass http://laoguang.me;  ##反向代理
                proxy_redirect off;
                proxy_set_header X-Real-IP $remote_addr;
 ##真实客户ip告诉后端
                proxy_set_header X-Forwarded-For Proxy_add_x_forwarded_for;
        }
        location /nginx {
                access_log off;
                stub_status on;  ##状态页面
        }
        error_page   500 502 503 504  /50x.html;
location = /50x.html {
            root   html;
        }
    }
}

4.2 拷贝到ng2上一份

scp /etc/nginx/nginx.conf 192.168.1.23:/etc/nginx/

4.3 测试反向代理能否负载均衡

lamp1,lamp2启动httpd

service httpd start

ng1重启nginx

service nginx restart

用RealIp访问测试能否轮询
http://192.168.1.22

同样测试ng2,如果都能实现负载均衡，那么继续

五.测试keepalived与nginx配合运行

现在192.168.1.18在 ng2上，        访问 http://192.168.1.18 测试能否轮询
ng2上 service keepalived stop     访问测试 http://192.168.1.18 能否轮询
关闭lamp1上的service httpd stop   访问测试http://192.168.1.18 是否会报错

到此高可用webserver构建完毕，没有单点故障，任何一点故障不影响业务。