求教一个LVS的问题

magicway

主备两台LVS服务器，后边两台Real server。用的直接路由的模式

测试了一阵发现有时VIP会同时飘到两台机器上。造成应用服务器不能访问。但关掉一台lvs服务器或者另外一台服务器，vip和服务都能切到ACTIVE的一端，但过一段时间，发现两台LVS服务器又同时起了。停掉LVS2的服务以后，vip飘到LVS1上，应用才能访问。刚开始接触lvs，请教一下可能什么地方的问题？
LVS1

serial_no = 245
primary = 192.168.111.11
service = lvs
backup_active = 1
backup = 192.168.111.12
heartbeat = 1
heartbeat_port = 1050
keepalive = 10
deadtime = 18
network = direct
nat_nmask = 255.255.255.255
debug_level = NONE
monitor_links = 1
virtual stweb {
     active = 1
     address = 192.168.111.100 bond0:1
     vip_nmask = 255.255.255.0
     port = 80
     persistent = 5
     pmask = 255.255.255.255
     send = "GET / HTTP/1.1\r\n\r\n"
     expect = "HTTP"
     use_regex = 0
     load_monitor = rup
     scheduler = wlc
     protocol = tcp
     timeout = 20
     reentry = 15
     quiesce_server = 1
     server app1 {
         address = 192.168.111.21
         active = 1
         weight = 10
     }
     server app2 {
         address = 192.168.111.22
         active = 1
         weight = 10
     }
}
virtual stkiosk {
     active = 1
     address = 192.168.111.200 bond0:2
     vip_nmask = 255.255.255.0
     port = 88
     send = "GET /stsf/services/Version HTTP/1.1\r\nHost: 192.168.111.200:88\r\n\r\n"
     expect = "HTTP/1.1 200"
     use_regex = 0
     load_monitor = rup
     scheduler = wlc
     protocol = tcp
     timeout = 20
     reentry = 15
     quiesce_server = 1
     server app1 {
         address = 192.168.111.21
         active = 1
         weight = 5
     }
     server app2 {
         address = 192.168.111.22
         active = 1
         weight = 5
     }
}

/var/log/message:

Nov 17 21:24:05 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:24:10 lvs1 pulse[29232]: gratuitous lvs arps finished
Nov 17 21:24:15 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:24:20 lvs1 pulse[29257]: gratuitous lvs arps finished
Nov 17 21:24:25 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:24:30 lvs1 pulse[29274]: gratuitous lvs arps finished
Nov 17 21:24:35 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:24:40 lvs1 pulse[29299]: gratuitous lvs arps finished
Nov 17 21:24:45 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:24:50 lvs1 pulse[29316]: gratuitous lvs arps finished
Nov 17 21:24:55 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:25:00 lvs1 pulse[29343]: gratuitous lvs arps finished
Nov 17 21:25:05 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:25:10 lvs1 pulse[29360]: gratuitous lvs arps finished
Nov 17 21:25:15 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:25:20 lvs1 pulse[29385]: gratuitous lvs arps finished
Nov 17 21:25:25 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:25:30 lvs1 pulse[29402]: gratuitous lvs arps finished
Nov 17 21:25:35 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:25:40 lvs1 pulse[29427]: gratuitous lvs arps finished
Nov 17 21:25:45 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:25:50 lvs1 pulse[29444]: gratuitous lvs arps finished
Nov 17 21:25:55 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:26:00 lvs1 pulse[29469]: gratuitous lvs arps finished
Nov 17 21:26:05 lvs1 pulse[3772]: partner active: resending arps
Nov 17 21:26:10 lvs1 pulse[29487]: gratuitous lvs arps finished

lvs2

serial_no = 247
primary = 192.168.111.12
service = lvs
backup_active = 1
backup = 192.168.111.11
heartbeat = 1
heartbeat_port = 1050
keepalive = 10
deadtime = 18
network = direct
nat_nmask = 255.255.255.255
debug_level = NONE
monitor_links = 1
virtual stweb {
     active = 1
     address = 192.168.111.100 bond0:1
     vip_nmask = 255.255.255.0
     port = 80
     persistent = 5
     pmask = 255.255.255.255
     send = "GET / HTTP/1.1\r\n\r\n"
     expect = "HTTP"
     use_regex = 0
     load_monitor = rup
     scheduler = wlc
     protocol = tcp
     timeout = 20
     reentry = 15
     quiesce_server = 1
     server app1 {
         address = 192.168.111.21
         active = 1
         weight = 10
     }
     server app2 {
         address = 192.168.111.22
         active = 1
         weight = 10
     }
}
virtual stkiosk {
     active = 1
     address = 192.168.111.200 bond0:2
     vip_nmask = 255.255.255.0
     port = 88
     send = "GET /stsf/services/Version HTTP/1.1\r\nHost: 192.168.111.200:88\r\n\r\n"
     expect = "HTTP/1.1 200"
     use_regex = 0
     load_monitor = rup
     scheduler = wlc
     protocol = tcp
     timeout = 20
     reentry = 15
     quiesce_server = 1
     server app1 {
         address = 192.168.111.21
         active = 1
         weight = 5
     }
     server app2 {
         address = 192.168.111.22
         active = 1
         weight = 5
     }
}

/var/log/message:

Nov 17 21:24:49 lvs2 nanny[3447]: The following exited abnormally:
Nov 17 21:24:49 lvs2 nanny[3447]: failed to read remote load
Nov 17 21:24:50 lvs2 nanny[3446]: The following exited abnormally:
Nov 17 21:24:50 lvs2 nanny[3446]: failed to read remote load
Nov 17 21:24:50 lvs2 pulse[2883]: partner active: resending arps
Nov 17 21:24:52 lvs2 nanny[3449]: The following exited abnormally:
Nov 17 21:24:52 lvs2 nanny[3449]: failed to read remote load
Nov 17 21:24:55 lvs2 pulse[28569]: gratuitous lvs arps finished
Nov 17 21:25:00 lvs2 pulse[2883]: partner active: resending arps
Nov 17 21:25:05 lvs2 nanny[3450]: The following exited abnormally:
Nov 17 21:25:05 lvs2 nanny[3450]: failed to read remote load
Nov 17 21:25:05 lvs2 pulse[28587]: gratuitous lvs arps finished
Nov 17 21:25:09 lvs2 nanny[3447]: The following exited abnormally:
Nov 17 21:25:09 lvs2 nanny[3447]: failed to read remote load
Nov 17 21:25:10 lvs2 nanny[3446]: The following exited abnormally:
Nov 17 21:25:10 lvs2 nanny[3446]: failed to read remote load
Nov 17 21:25:10 lvs2 pulse[2883]: partner active: resending arps
Nov 17 21:25:12 lvs2 nanny[3449]: The following exited abnormally:
Nov 17 21:25:12 lvs2 nanny[3449]: failed to read remote load
Nov 17 21:25:15 lvs2 pulse[28607]: gratuitous lvs arps finished
Nov 17 21:25:20 lvs2 pulse[2883]: partner active: resending arps
Nov 17 21:25:25 lvs2 nanny[3450]: The following exited abnormally:
Nov 17 21:25:25 lvs2 nanny[3450]: failed to read remote load
Nov 17 21:25:25 lvs2 pulse[28626]: gratuitous lvs arps finished
Nov 17 21:25:26 lvs2 pulse[2883]: Terminating due to signal 15
Nov 17 21:25:26 lvs2 lvs[3434]: shutting down due to signal 15
Nov 17 21:25:26 lvs2 lvs[3434]: shutting down virtual service stweb
Nov 17 21:25:26 lvs2 nanny[3446]: Terminating due to signal 15
Nov 17 21:25:26 lvs2 nanny[3447]: Terminating due to signal 15
Nov 17 21:25:26 lvs2 lvs[3434]: shutting down virtual service stkiosk
Nov 17 21:25:26 lvs2 nanny[3449]: Terminating due to signal 15
Nov 17 21:25:26 lvs2 nanny[3450]: Terminating due to signal 15

kns1024wh

测试了一阵发现有时VIP会同时飘到两台机器上。造成应用服务器不能访问。但关掉一台lvs服务器或者另外一台服务器，vip和服务都能切到ACTIVE的一端，但过一段时间，发现两台LVS服务器又同时起了。停掉LVS2的服务以后，vip飘到LVS1上，应用才能访问

这种现象说明，你使用的lvs支持高可用性的组件的心跳信息是检测失败的；LVS本身作为一个节点使用的时候会有单点故障，所以就需要正好Heartbeat，Heartbeat保证lvs调度节点的高可用，同事还要结合第三方组件对realserver的状态进行监控，将失效的节点从lvs路由调度信息中移除

冷风

心跳断了，检查2个机器之间的网络心跳通讯