LVS HA切换的时候发送大量IPVS数据包，导致组播风暴

环境 v1.24 keepalive 1.19 dr模式
配置（VRRP两边配置都是BACKUP）
vrrp_instance VI_1 {
state BACKUP
interface eth0
lvs_sync_daemon_interface eth0
virtual_router_id 51
priority 150
advert_int 1
smtp_alert
nopreempt

authentication {
auth_type PASS
auth_pass
}

virtual_ipaddress {
192.168.23 dev eth0 label eth0:1
}
}

virtual_server IP 端口 {

delay_loop 8
lb_algo lc
lb_kind DR
persistence_timeout 30
protocol TCP

real_server IP 端口 {

weight 1
TCP_CHECK {
connect_port 端口
connect_timeout 50
}

}
}

在切换的时候 /var/log/message
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Reloading Healthchecker child process(21823) on signal
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Reloading VRRP child process(21825) on signal
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Netlink reflector reports IP 192.168.238.248 added
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Registering Kernel netlink reflector
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Registering Kernel netlink command channel
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Opening file '/usr/local/keepalived/etc/keepalived.conf'.
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Configuration is using : 168867 Bytes
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: service [192.168.239.210:3307] no longer exist
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Removing service [] from VS []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Using LinkWatch kernel netlink reflector...
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Netlink reflector reports IP 192.168.238.248 added
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Registering Kernel netlink reflector
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Registering Kernel netlink command channel
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Registering gratutious ARP shared channel
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Opening file '/usr/local/keepalived/etc/keepalived.conf'.
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Configuration is using : 70354 Bytes
Jul 22 16:24:11 db-lvs-238-248 kernel: IPVS: stopping sync thread 18400 ...
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Using LinkWatch kernel netlink reflector...
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: VRRP_Instance(VI_1) Entering BACKUP STATE
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: VRRP sockpool: [ifindex(2), proto(112), fd(16,17)]
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: Remote SMTP server [] connected.
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []
Jul 22 16:24:11 db-lvs-238-248 Keepalived_vrrp: SMTP alert successfully sent.
Jul 22 16:24:11 db-lvs-238-248 Keepalived_healthcheckers: Activating healtchecker for service []

捉包发现存在大量重复的IPVS数据包。

感觉就像进入死循环，最高IPVS数据包流量能达到几百M。形成组播风暴，影响其它服务。但自己的HA切换能正确进行。