LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: [lvs-users] Fwd: InActConn dilemma

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: [lvs-users] Fwd: InActConn dilemma
From: Andre Magri <andremag@xxxxxxxxx>
Date: Sun, 14 Feb 2010 11:58:48 +0100
That was a great explanation Graeme thank you so much.  Indeed enable
keepalives was disabled on www1.

I have one more question for you though :)
Wouldn't it be less taxing on the loadbalancer to have keepalives enabled?
At one point we had a slowdown on the responsiveness of the site and I
thought at the time that having all those inactive connections were a
contributing factor.

Cheers.

On 14 February 2010 10:40, Graeme Fowler <graeme@xxxxxxxxxxx> wrote:

> On Sun, 2010-02-14 at 10:16 +0100, Andre Magri wrote:
> > Our two servers are the exact same Dell Poweredge model with the same
> cpu's
> > and same memory modules except for the network cards where one of them is
> > running in a team.  They are running IIS and I can see from perfmon that
> one
> > of them (www1) is getting a high amount of requests where as the other is
> > quite free.
>
> OK, that's good to know. Now we can concentrate on...
>
> > If I observe connections on both servers (ipvsadm -lcn), connections to
> www2
> > are being kept established and therefore further traffic is done through
> > these already 'active' connections, listed as ESTABLISHED, until they are
> > timed out. www1 (or LVS, I'm getting confused here) is not keeping these
> > connections alive and therefore turn into FIN_WAIT state until the 2
> minutes
> > are up.
>
> This.
>
> > Which one is behaving incorrectly here - www1 or www2?  This will help me
> > establish where the problem is.
>
> They're both behaving correctly :)
>
> It looks as though you have IIS on www1 set to not do keepalives so
> every request *for every element on a page* is a separate connection.
>
> www2, however, seems to be honouring keepalive requests so is allowing a
> given client to make multiple HTTP requests down the same TCP
> connection.
>
> Both of these are perfectly normal behaviour and are dependent on
> circumstance, however it would be easier for you to handle the
> connections (in terms of raw numbers, nothing else) if the IIS server on
> both systems was configured in the same way.
>
> You might need to check the following:
>
> http://technet.microsoft.com/en-us/library/cc772183(WS.10).aspx<http://technet.microsoft.com/en-us/library/cc772183%28WS.10%29.aspx>
>
> HTH
>
> 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>