Re: [Xen-users] Arp not sent after live migrations on xen-unstable

[George Shuklin <george.shuklin@xxxxxxxxx>]

Good day.

I found that newer version of kernel (2.6.34-xen) from gentoo (wich one
is stealed somewere (suse? not sure)) adopted to debian works fine.

At least within my tests (4 cycles of migration between every of 6
testings servers) I was unable to reproduce bugs.

I'll continue to tests, if I was wrong and bug still exists, I'll mail
about this.

PS Migration tested on XCP 0.5 by xe vm-migrate command.

Ð ÐÑÐ, 19/08/2010 Ð 10:11 -0700, Nathan March ÐÐÑÐÑ:
> The consensus from the mailing list posts I found a while ago seemed 
> to imply that it had been fixed so I'm not clear if this is an 
> outstanding bug, a regression or just something that I'm doing wrong 
> that has broken it.
> 
> Are there any steps I can take to debug this further and try and get 
> some more info? Cranking up the debug level on xend didn't seem to 
> result in anything useful.
> 
> - Nathan
> 
> On 8/18/2010 5:53 PM, George Shuklin wrote:
> > I'm pretty concerned about this problem.
> >
> > Right now: I suppose, it was fixed somewhere in new netfront driver.
> > (not checked).
> >
> > If no, I have few ideas about workaround. Main is an event via xen-evtch
> > about suspending. Small module/appilcation in guest can sent a fake ARP
> > every time it got suspend event. Other is using xenstore to send command
> > to program from host. (both are ugly hacks).
> >
> > Best way is still fake ARP from netback or netfront...
> >
> > Ð ÐÑÐ, 18/08/2010 Ð 14:22 -0700, Nathan March ÐÐÑÐÑ:
> >> Hi All,
> >>
> >> Having an issue with xen-4.0-unstable where after completing a live
> >> migration, an arp is never sent out to update the switch with the new
> >> location of the mac address.
> >>
> >> Network is configured as per libvirt's instructions with the xen
> >> network-script disabled and the bridge (xenbr0) always up.
> >> http://wiki.libvirt.org/page/Networking#Bridged_networking_.28aka_.22shared_physical_device.22.29
> >>
> >> Things work fine 10s after doing the migration once the switch has
> >> picked up the move, checking the mac address tables confirms it's still
> >> going to the old host. I've also confirmed that it's not going out at
> >> all via a span port on the switch.
> >>
> >> Saw a bunch of postings a while ago regarding it, but nothing recent.
> >> Anyone have any insight on what might be wrong?
> >>
> >> - Nathan
> >>
> >> _______________________________________________
> >> Xen-users mailing list
> >> Xen-users@xxxxxxxxxxxxxxxxxxx
> >> http://lists.xensource.com/xen-users
> >
> >
> > _______________________________________________
> > Xen-users mailing list
> > Xen-users@xxxxxxxxxxxxxxxxxxx
> > http://lists.xensource.com/xen-users



_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users