LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Problem with LVS-DR and Windows NT

To: <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Problem with LVS-DR and Windows NT
From: Thierry Coopman <calvin@xxxxxxxxx>
Date: Thu, 07 Mar 2002 13:28:13 +0100
On 07-03-2002 11:17, "Joseph Mack" <mack.joseph@xxxxxxx> wrote:

> Thierry Coopman wrote:
>> 
>> On 06-03-2002 16:54, "Joseph Mack" <mack.joseph@xxxxxxx> wrote:
>> 
>>> Thierry Coopman wrote:
> 
>>> What is the 'real ip'? It's not the IP of the default gw?
>> 
>> That's what I've put in, so that the real ip can send out the packet to the
>> network. This seems to work a lot better :)
> 
> thanks, but which machine is it the IP of, the default gw?
> 
> Joe
> 

The machine is on a private class C (192.168.202.x), the default gateway is
192.168.202.1 and the real ip is 192.168.202.141. The VIP is 192.168.202.140
witch is attributed to the MS loopback adapter. This adapter has a default
gateway 192.168.202.141 and I put the mac  of that ip in the static arp
table.

So the way I see it, is that the MS loopback adapter cannot send anything on
the LAN (wich is great since that's what we want to avoid it answering to
arp requests or sending gratious arps of its VIP).

If you want to send data from this interface, it's possible to either leave
the gateway blank (so the 'default' route applies I guess) or to specify the
real-ip of the server. The packets will get passed to the real-ip (since
it's on the same machine and the mac is known as a static arp entry, it's
possible), witch in turn will send them on the LAN network.

Anyway, that's the way I guess things are going on NT. I'm more a fan of
*nix systems that have a more straight forward TCP/IP stack approach.
Microsoft invented a proper stack for NT 4.0, and I really don't want to
know all particularities :))



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