[ovs-discuss] Fwd: Problems with WinXP vms

Tom Brown nextstate at gmail.com
Wed Apr 13 01:09:36 UTC 2011

On Tue, Apr 12, 2011 at 11:38, Ben Pfaff <blp at nicira.com> wrote:

> On Mon, Apr 11, 2011 at 01:39:01PM -0700, Tom Brown wrote:
> > The version of ovs I'm using is: v1.1.0pre2 - 13 Sep 2010
> Pretty old.  Can you upgrade to version v1.1.0 that was just released
> last week?
> > I believe I solved the issue with not being able to access some of the
> vms.
> > I removed an erroneous default route and all the vms were accessible.
> Glad to hear that.
> > There is something to note about our application. We create additional
> > mac/ip pairs on the XP vms. If we create one mac/ip pair on the XP vm,
> the
> > application succeeds without errors. If we create two or more mac/ip
> pairs,
> > we get the errno 87 every time we perform a "load" from the client to the
> > server. The initial connection always works, but the "load" is when the
> > mac/ip pairs are generated. Additionally, the ip is provided through
> > However, when the ip is allocated statically, the "load" works without
> > error.
> What is the externally visible effect of creating a mac/ip pair in an
> XP VM?  Does another network device (vif, tap, whatever) get created
> in the host?  If so, is that network device getting added to the
> vswitch's bridge?  What does the DHCP traffic flowing back and forth
> across the new network device look like?

Thank you very much for your questions! They led me to the answer. I did
upgrade to v1.1.0 and it helped out a little. However, your questions about
dhcp made me scrutinize my dhcp server setup. I decided to change my setup
to one network only so I could simplify the dhcpd.conf considerably. The end
result was I could perform a "load" that created many mac/ip pairs without a
problem. I think the solution for several networks is to spawn a dhcp daemon
for each network.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20110412/49dbbe17/attachment-0001.html>

More information about the discuss mailing list