LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: [lvs-users] ldirectord unbalance load distribution

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: [lvs-users] ldirectord unbalance load distribution
From: Liu Yan <gzbigegg.ml@xxxxxxxxx>
Date: Mon, 28 Sep 2009 00:35:19 +0800
Thanks Graeme, for the reply.
Here's my result:

1) ipvsadm -Ln --stats
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port               Conns   InPkts  OutPkts  InBytes
OutBytes
  -> RemoteAddress:Port
TCP  VIP1:80              1207524 12718506        0    2502M        0
  -> RIP1-2:80                48784   456988        0 68388231        0
  -> RIP1-1:80                719206  7807994        0    1451M        0
TCP  VIP2:80               2631485 27031952        0    5674M        0
  -> RIP2-2:80               1082721 10626928        0    2422M        0
  -> 1RIP2-1:80               1548764 16405024        0    3251M        0

2) ipvsadm -Ln --rate
IP Virtual Server version 1.2.1 (size=4096)
Prot LocalAddress:Port                 CPS    InPPS   OutPPS    InBPS
OutBPS
  -> RemoteAddress:Port
TCP  VIP1:80                    3       28        0     5768        0
  -> RIP1-2:80                    0        0        0        0        0
  -> RIP1-1:80                     3       28        0     5768        0
TCP  VIP2:80                     7       71        0    18218        0
  -> RIP2-2:80                     0        0        0        0        0
  -> RIP2-1:80                     7       71        0    18218        0

Questions:

-- in "--stats" command output, the number 68388231 doesn't make any sense
to me, what's it?
-- the statistical info is reset on each restart, is it true? if it is true,
the load seems not balance (I am using the lc scheduler). Why?
-- What are those numbers representing in the "--rate" command output? I am
seeing all 0 for the remote server, why?

Thanks,
Yan

2009/9/27 Graeme Fowler <graeme@xxxxxxxxxxx>

> On Sun, 2009-09-27 at 02:56 +0800, Liu Yan wrote:
> > 2) Why the remote (Route) server is taking much more requests than the
> local
> > one?
>
> It isn't.
>
> > 3) Why the local server has a large number of "InActConn"?
>
> Because it should.
>
> To explain: in DR mode the director does not see the entire TCP
> transaction (it only sees one side, the client side) since the
> realservers respond directly to the client.
>
> When one of the realservers is "local" then full state can be kept, so
> the director knows how many active (ie. open in TCP terms) and inactive
> (ie. in TIME_WAIT or FIN_WAIT state) connections exist. This looks to
> the inexperienced eye as a real imbalance in load distribution but it
> may not be the case in reality.
>
> You'd be much better off using the other statistical options to ipvsadm:
>
> ipvsadm -Ln --stats
> ipvsadm -Ln --rate
>
> These will give you a much better view of how many actual connections
> you're handling.
>
> Graeme
>
>
> _______________________________________________
> Please read the documentation before posting - it's available at:
> http://www.linuxvirtualserver.org/
>
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://lists.graemef.net/mailman/listinfo/lvs-users
>
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/

LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users

<Prev in Thread] Current Thread [Next in Thread>