RedHat AS 3.0 LVS Director quits randmomly

[Adi Linden <adil@xxxxxxxxxx>]

Hi,

I have two LVS-NAT directors (RedHat AS 3.0 w/kernel-smp-2.4.21-15.0.2.EL 
kernel) in a failover configuration. There are a variety of real servers 
behind these directors. 

The problem that occurs is that the static nat explained below ceases to 
work after a few hours. It seems the only way to restore proper operation 
is to reboot the director.  

One of these real servers is a mail server. When the mail server initiates 
outgoing connection it needs to use the VIP assigned to it rather than the 
ip of the director that NAT on the director defaults to. The RIP of the 
mail server is 172.28.1.25 and the VIP is 66.165.220.47 for outgoing 
connections and inbound ssh and another VIP is 66.165.220.14 for smtp. The 
66.165.220.14 is directed to another real server for pop3 and imap access. 
In order to accomplish the static NAT of 172.28.1.25 to 66.165.220.47 I 
added the following iptables rule on the directors:

    iptables -v -t nat -I POSTROUTING -s 172.28.1.25 \
        -j SNAT --to-source 66.165.220.47

The /etc/sysconfig/ha/lvs.cf file has the following entries regarding the 
services of the mail server:

virtual mailsmtp.knet.ca_ssh {
     active = 1
     address = 66.165.220.47 eth0:7
     vip_nmask = 255.255.255.192
     port = 22
     use_regex = 0
     load_monitor = none
     scheduler = wlc
     protocol = tcp
     timeout = 6
     reentry = 15
     quiesce_server = 0
     server mailsmtp1.knet.ca {
         address = 172.28.1.25
         active = 1
         weight = 1
     }
}
virtual mail.knet.ca_smtp {
     active = 1
     address = 66.165.220.14 eth0:8
     vip_nmask = 255.255.255.192
     port = 25
     use_regex = 0
     load_monitor = none
     scheduler = wlc
     protocol = tcp
     timeout = 90
     reentry = 120
     quiesce_server = 0
     server mailsmtp1.knet.ca {
         address = 172.28.1.25
         active = 1
         weight = 1
     }
}
 
Thanks,
Adi