Re: lvs not setting up real servers

On Sun, 2 Mar 2003, Greg Boehnlein wrote:

> Hello,
> 	I have a pair of RH AS 2.1 LVS Routers in direct routing 
> mode configured as the front end for two back-end services. The goal of 
> this cluster is to provide HA Load Balancing to a pair of backend 
> Mail/Web/Ftp servers. The problem I am having is that the "lvs" command 
> does not seem to be adding all of the rules defined in the lvs.cf file. 
> Specifically, it seems to skip the pair of real-servers that I have 
> defined for my SMTP cluster. I have verified this behavior by adding the 
> ipvsadm commands by hand to get the services to work. When I do this, it 
> works fine. But I can't get the system to do it auomatically either by 
> using the Piranha GUI or by edting the lvs.cf by hand.

Ignore all this. The problem was very simple that I was using the default 
Expect/Send strings for a Webserver for all of the services. As a result, 
Nanny was failing when it did it's initial checks and not bringing up the 
real-servers for that service.
 
> Here is the lvs.cf file I am using:
> 
> serial_no = 61
> primary = 207.166.192.17
> service = lvs
> backup_active = 1
> backup = 207.166.192.18
> heartbeat = 1
> heartbeat_port = 539
> keepalive = 6
> deadtime = 18
> network = direct
> reservation_conflict_action = preempt
> debug_level = 9
> virtual mail.nacs.net {
>      active = 1
>      address = 207.166.192.19 eth0:1
>      port = 25
>      send = "quit\r\n"
>      expect = "221"
>      load_monitor = none
>      scheduler = wlc
>      protocol = tcp
>      timeout = 6
>      reentry = 15
>      quiesce_server = 0
>      server node2.nacs.net {
>          address = 207.166.192.34
>          active = 1
>          weight = 1
>      }
>      server tracy.nacs.net {
>          address = 207.166.192.36
>          active = 1
>          weight = 1
>      }
> }
> 
> virtual www.nacs.net {
>      active = 1
>      address = 207.166.192.19 eth0:1
>      port = 80
>      send = "GET / HTTP/1.0\r\n\r\n"
>      expect = "HTTP"
>      load_monitor = none
>      scheduler = wlc
>      protocol = tcp
>      timeout = 6
>      reentry = 15
>      quiesce_server = 0
>      server node2.nacs.net {
>          address = 207.166.192.34
>          active = 1
>          weight = 1
>      }
>      server tracy.nacs.net {
>          address = 207.166.192.36
>          active = 1
>          weight = 1
>      }
> 
> 
> When "pulse" is started, the ipvsadm command returns the following:
> 
> IP Virtual Server version 0.8.1 (size=65536)                   
> Prot LocalAddress:Port Scheduler Flags                         
>   -> RemoteAddress:Port             Forward Weight ActiveConn InActConn
> TCP  cluster.nacs.net:smtp wlc
> TCP  cluster.nacs.net:http wlc
>   -> tracy.nacs.net:http            Route   1      0          0         
>   -> node2.nacs.net:http            Route   1      0          0         
> 
> I've tracked the particular program responsible for adding the 
> ipvsadm commands to "lvs". An Strace of "lvs" shows the following calls to 
> ipvsadm:
> 
> 4731  execve("/sbin/ipvsadm", ["/sbin/ipvsadm", "-C"], [/* 23 vars */]) = 0
> 
> 4734  execve("/sbin/ipvsadm", ["/sbin/ipvsadm", "-A", "-t", 
> "207.166.192.19:80", "-s", "wlc"], [/* 23 vars */]) = 0
> 
> 4737  execve("/sbin/ipvsadm", ["/sbin/ipvsadm", "-A", "-t", 
> "207.166.192.19:25", "-s", "wlc"], [/* 23 vars */] <unfinished>
> 
> 4740  execve("/sbin/ipvsadm", ["/sbin/ipvsadm", "-a", "-t", 
> "207.166.192.19:80", "-r", "207.166.192.36", "-g", "-w", "1"], [/
> * 23 vars */] <unfinished ...>
> 
> 4741  execve("/sbin/ipvsadm", ["/sbin/ipvsadm", "-a", "-t", 
> "207.166.192.19:80", "-r", "207.166.192.34", "-g", "-w", "1"], [/
> * 23 vars */] <unfinished ...>
> 
> I can see no commands being executed to add the real servers for the SMTP 
> clustering. Hence, the clustering does not work. This is bad.
> 
> So, I added the rules by hand:
> 
> [root@atm1 ha]# cat ipvsadm-rules 
> -A -t cluster.nacs.net:smtp -s wlc
> -a -t cluster.nacs.net:smtp -r tracy.nacs.net:smtp -g -w 1
> -a -t cluster.nacs.net:smtp -r node2.nacs.net:smtp -g -w 1
> -A -t cluster.nacs.net:http -s wlc
> -a -t cluster.nacs.net:http -r tracy.nacs.net:http -g -w 1
> -a -t cluster.nacs.net:http -r node2.nacs.net:http -g -w 1
> 
> [root@atm1 ha]# ipvsadm -R < ipvsadm-rules 
> [root@atm1 ha]# ipvsadm
> IP Virtual Server version 0.8.1 (size=65536)                   
> Prot LocalAddress:Port Scheduler Flags                         
>   -> RemoteAddress:Port             Forward Weight ActiveConn InActConn
> TCP  cluster.nacs.net:smtp wlc
>   -> node2.nacs.net:smtp            Route   1      0          0         
>   -> tracy.nacs.net:smtp            Route   1      0          0         
> TCP  cluster.nacs.net:http wlc
>   -> node2.nacs.net:http            Route   1      0          0         
>   -> tracy.nacs.net:http            Route   1      0          0         
> 
> And everything works great. I guess my question is does anyone have a clue 
> as to what might be going on?
> 
> Thanks,
> Greg
> 
> 

-- 
    Vice President of N2Net, a New Age Consulting Service, Inc. Company
         http://www.n2net.net Where everything clicks into place!
                             KP-216-121-ST