在之前的篇幅中对LVS及keepalived都分别做了介绍，本篇开始总结下LVS+keepalived组合方案。这里以LVS的DR模式为例，在LB层再实现HA的功能 。具体加架如下图：

lvs-keepalived-master-backup
lvs-keepalived-master-backup

一、IP及规划

主机 IP地址
realserver 192.168.122.10
realserver 192.168.122.20
director 192.168.122.30
director 192.168.122.40
VIP 192.168.122.100

两台realserver安装apache httpd（yum -y install httpd）

两台director主机安装ipvsadm 、keepalived（yum -y install ipvsadm keepalived）

VIP需要配置在两台director上的eth0网口上，同时需要将该IP配置在两台realserver的lo回环口上。

在之前的篇幅中也介绍过，keepalived底层有关于IPVS的功能模块，可以直接在其配置文件中实现LVS的配置，不需要通过ipvsadm命令再单独配置。

二、director主机配置

MASTER节点的keepalived配置文件如下：

 1# cat /etc/keepalived/keepalived.conf
 2global_defs {
 3router_id LVS_T1
 4}
 5vrrp_sync_group bl_group {
 6group {
 7  bl_one
 8}
 9}
10vrrp_instance bl_one {
11    state MASTER
12    interface eth0
13    lvs_sync_daemon_interface eth0
14    virtual_router_id 38
15    priority 150
16    advert_int 3
17    authentication {
18        auth_type PASS
19        auth_pass 1111
20    }
21    virtual_ipaddress {
22      192.168.122.100
23    }
24}
25virtual_server 192.168.122.100 80 {
26    delay_loop 3
27    lb_algo rr
28    lb_kind DR
29    persistence_timeout 1
30    protocol TCP
31    real_server 192.168.122.10 80 {
32        weight 1
33        TCP_CHECK {
34            connect_timeout 10
35            nb_get_retry 3
36            delay_before_retry 3
37            connect_port 80
38        }
39    }
40    real_server 192.168.122.20 80 {
41        weight 1
42        TCP_CHECK {
43            connect_timeout 10
44            nb_get_retry 3
45            delay_before_retry 3
46            connect_port 80
47        }
48    }
49}

BACKUP director主机的配置文件如下：

 1# cat /etc/keepalived/keepalived.conf
 2global_defs {
 3router_id LVS_T1
 4}
 5vrrp_sync_group bl_group {
 6group {
 7  bl_one
 8}
 9}
10vrrp_instance bl_one {
11    state MASTER
12    interface eth0
13    lvs_sync_daemon_interface eth0
14    virtual_router_id 38
15    priority 150
16    advert_int 3
17    authentication {
18        auth_type PASS
19        auth_pass 1111
20    }
21    virtual_ipaddress {
22      192.168.122.100
23    }
24}
25virtual_server 192.168.122.100 80 {
26    delay_loop 3
27    lb_algo rr
28    lb_kind DR
29    persistence_timeout 1
30    protocol TCP
31    real_server 192.168.122.10 80 {
32        weight 1
33        TCP_CHECK {
34            connect_timeout 10
35            nb_get_retry 3
36            delay_before_retry 3
37            connect_port 80
38        }
39    }
40    real_server 192.168.122.20 80 {
41        weight 1
42        TCP_CHECK {
43            connect_timeout 10
44            nb_get_retry 3
45            delay_before_retry 3
46            connect_port 80
47        }
48    }
49}

从上面的配置可以看到，这里配置的健康检测方式是基于IP端口的，我们也可憎修改成基于URL的。这点可以参考 keepalived健康检查方式 。主机在启动keepalived服务时，可以从message中看到如下日志：

keepalived-message
keepalived-message

三、realserver主机配置

两台realserver上使用的脚本一样，内容如下：

 1# cat dr_client.sh
 2#!/bin/bash
 3VIP=192.168.122.100
 4BROADCAST=192.168.122.255  #vip's broadcast
 5. /etc/rc.d/init.d/functions
 6case "$1" in
 7 start)
 8  echo "reparing for Real Server"
 9    echo "1" >/proc/sys/net/ipv4/conf/lo/arp_ignore
10    echo "2" >/proc/sys/net/ipv4/conf/lo/arp_announce
11    echo "1" >/proc/sys/net/ipv4/conf/all/arp_ignore
12    echo "2" >/proc/sys/net/ipv4/conf/all/arp_announce
13    ifconfig lo:0 $VIP netmask 255.255.255.255 broadcast $BROADCAST up
14     /sbin/route add -host $VIP dev lo:0
15     ;;
16 stop)
17     ifconfig lo:0 down
18    echo "0" >/proc/sys/net/ipv4/conf/lo/arp_ignore
19    echo "0" >/proc/sys/net/ipv4/conf/lo/arp_announce
20    echo "0" >/proc/sys/net/ipv4/conf/all/arp_ignore
21    echo "0" >/proc/sys/net/ipv4/conf/all/arp_announce
22     ;;
23 *)
24     echo "Usage: lvs {start|stop}"
25     exit 1
26esac

四、测试

分别在两台主机上执行如下命令启动：

1# /etc/init.d/keepalived start
2# sh dr_client.sh start

在两台realserver上启动httpd服务，关闭防火墙（或放行80端口）。可以使用如下脚本进行访问测试：

1#!/bin/sh
2for((i=1;i<=100;i++));do
3  curl http://192.168.122.100 >> /tmp/q;
4done

1、简单测试

在任一台director主机上使用ipvsadm命令观察：

1[root@lvs-dr ~]# ipvsadm -Ln
2IP Virtual Server version 1.2.1 (size=4096)
3Prot LocalAddress:Port Scheduler Flags
4  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
5TCP  192.168.122.100:80 rr persistent 1
6  -> 192.168.122.10:80            Route   1      0          0
7  -> 192.168.122.20:80            Route   1      0          100

注：这里会现一个问题，发现使用轮询算法，使用一台client得到的结果是不均衡的（在一定的时间内其一直走其中一台realserver） 。当使用多台client去访问时，发现访问的结果是均衡的。

2、realserver故障测试

将其中一台realserver主机关闭，通过curl 查看发现其会每次请求获取的都是正常的主机的页面，message中也能看到如下日志：

1Oct  7 22:35:10 lvs-dr Keepalived_vrrp[32634]: IPVS: Daemon has already run
2Oct  7 22:35:10 lvs-dr Keepalived_healthcheckers[32633]: Netlink reflector reports IP 192.168.122.100 added
3Oct  7 22:35:15 lvs-dr Keepalived_vrrp[32634]: VRRP_Instance(bl_one) Sending gratuitous ARPs on eth0 for 192.168.122.100
4Oct  7 22:38:16 lvs-dr Keepalived_healthcheckers[32633]: TCP connection to [192.168.122.10]:80 failed !!!
5Oct  7 22:38:16 lvs-dr Keepalived_healthcheckers[32633]: Removing service [192.168.122.10]:80 from VS [192.168.122.100]:80

这台realserver恢复后，message中的日志如下：

1Oct  7 22:39:19 lvs-dr Keepalived_healthcheckers[32633]: TCP connection to [192.168.122.10]:80 success.
2Oct  7 22:39:19 lvs-dr Keepalived_healthcheckers[32633]: Adding service [192.168.122.10]:80 to VS [192.168.122.100]:80

3、director主机故障测试

通过关闭master director主机的keepalived服务，在backup director主机上发现日志如下：

1Oct  7 22:40:14 lvs-dr Keepalived_vrrp[1601]: VRRP_Instance(bl_one) Transition to MASTER STATE
2Oct  7 22:40:14 lvs-dr Keepalived_vrrp[1601]: VRRP_Group(bl_group) Syncing instances to MASTER state
3Oct  7 22:40:17 lvs-dr Keepalived_vrrp[1601]: VRRP_Instance(bl_one) Entering MASTER STATE
4Oct  7 22:40:17 lvs-dr Keepalived_vrrp[1601]: VRRP_Instance(bl_one) setting protocol VIPs.
5Oct  7 22:40:17 lvs-dr Keepalived_vrrp[1601]: VRRP_Instance(bl_one) Sending gratuitous ARPs on eth0 for 192.168.122.100
6Oct  7 22:40:17 lvs-dr Keepalived_healthcheckers[1600]: Netlink reflector reports IP 192.168.122.100 added

此时通过122.100访问服务不受影响 。