LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Ultramonkey/LVS-TUN connection stalls on client

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Ultramonkey/LVS-TUN connection stalls on client
From: Horms <horms@xxxxxxxxxxxx>
Date: Thu, 26 Feb 2004 18:02:18 +0900
On Thu, Feb 26, 2004 at 02:47:27AM -0600, Nigel Hamilton wrote:
> > > >         But I've followed the UltraMonkey installation instructions 
> > > > word-for-word and there has been no mention of a tunl0 device. Is 
> > > > tunl0 the equivalent of lo:0 in the UltraMonkey docs?
> > > 
> > > Looks like Horms gets to update his docs, thanks for being his guinea pig
> > > :).  (BTW patches for documentation accepted..)
> > 
> > Ultra Monkey makes no reference on using LVS-TUN, it is not incuded in
> > the documentation, and at this stage I have no plans to add it. If you
> > would like to write documentation for a new topology that includes
> > LVS-TUN I would consider it. But frankly, Ultra Monkey is aimed at load
> > balancing on a LAN, I am yet to see a situation where LVS-TUN is needed
> > for that.
> > 
> 
> HI Horms,
> 
>       I think it would be great for the ultramonkey newbie to see this 
> mentioned somewhere.

Hi Nigel,

No problem, I am happy to clarify the documentation. Do
you have some ideas on where this should be done?

>       One of the reasons I went with Ultramonkey was a glowing report 
> from one of the users at my ISP - balancing may need to go across more 
> than one LAN in multiple data centers. I assumed they used TUN.

In that case you probably do need LVS-TUN, unfortunately that is
not something that the Ultra Monkey documentation addresses.
As I mentioned above I would be willing to consider a contributed
topology for this, but it isn't a priority for me right now.

>       As it turns out I'm still trying to get this to work. When I added 
> an extra RealServer the connection to the first RealServer died.

Boo. Certainly this should not occur. Do you know if LVS mysteriously
tried to send the connection elsewhere or if it just dropped the
connection? Playing with /proc/sys/net/ipv4/vs/debug_level may yield
some aditional information, but tcpdump is probably the best source of
debugging information.

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