Re: Large HTTP GET/POST revisited (and solved)

[Graham Purcocks <grahamp@xxxxxxxxxxxxx>]

Mack.Joseph@xxxxxxxxxxxxxxx wrote:
> Joseph Mack PhD, High Performance Computing & Scientific Visualisation
> LMIT, Supporting the EPA Research Triangle Park, NC 919-541-0007
> Federal Contact - John B. Smith 919-541-1087 - smith.john@xxxxxxx
>
> lvs-users-bounces+mack.joseph=epamail.epa.gov@xxxxxxxxxxxxxxxxxxxxxx
> wrote on 03/11/2005 11:13:12 AM:
>
>
> > > can you give me a reference? Which 2.6.x kernels have it?
> > I honestly have no idea.
>
> there's not a word about it in the 2.6.5 Documentation directories
> (grep'ed for mtu, tunl, tunnel).
>
>
>
> > I have LVS setup on a version of 2.4.20 and had to change
> > the MTU values to prevent the problem.
> >
> > When I was investigating, I was told (by a non LVS bod)
> > there is a bug in the IP Encapsulation code in most
> > versions of the 2.4 kernel where by when negotiating the
> > MTU for the IP tunnel it does not take the overhead of the
> > tunnel into account.
>
> this sounds like what we've got.
>
> Still someone had to work out how to handle it in windows too.
Not necesarily. I had similar problems with VPN tunnels and Windows XP 
SP2 in that it didn't negotiate the MTU correctly when I upgraded to 
ADSL. I had to change the MTU in the VPN virtual IP stack. There were 
some tools to work out the correct value. So Mr Gates may well have 
exactly the same problem.
> > It is very difficult to reproduce. The bug only occurs if
> > the packet is a very specific size. From my testing, even
> > source and destination IP address makes a difference.
>
> let's see how Casey did it
>
> Joe
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://www.in-addr.de/mailman/listinfo/lvs-users