[ovirt-users] Network Interface order changed after reboot

Dan Kenigsberg danken at redhat.com
Mon Jun 27 06:29:48 UTC 2016


On Sun, Jun 26, 2016 at 03:59:31PM +0300, Yedidyah Bar David wrote:
> On Sun, Jun 26, 2016 at 3:51 PM, Edward Haas <ehaas at redhat.com> wrote:
> >
> >
> > On Thu, Jun 23, 2016 at 3:49 PM, <ovirt at timmi.org> wrote:
> >>
> >> Hi List,
> >>
> >> I have two nodes (running CentOS 7) and the network interface order
> >> changed for some interfaces after every reboot.
> >>
> >> The configurations are done through the oVirt GUI. So the ifcfg-ethX
> >> scripts are configured automatically by VDSM.
> >>
> >> Is there any option to get this configured to be stable?
> >>
> >> Best regards and thank you
> >>
> >> Christoph
> >>
> >> _______________________________________________
> >> Users mailing list
> >> Users at ovirt.org
> >> http://lists.ovirt.org/mailman/listinfo/users
> >
> >
> > Hi Christoph,
> >
> > VDSM indeed edits and takes ownership of the interfaces for the networks it
> > manages.
> > However, editing the ifcfg files should not change anything in the order of
> > the devices, unless it was originally set
> > in an unsupported fashion. An ifcfg file is bound to a specific device name
> > and I'm not familiar to device names
> > floating around randomly.
> > Perhaps you should elaborate more on what it means by 'order changed'.
> >
> > Here is an example of a setup we do not support (pre adding the host to
> > Engine):
> > The initial ifcfg file name: ifcfg-eth0
> > The initial ifcfg file content: DEVICE="eth1"
> > In this configuration, the name of the ifcfg file is inconsistent with the
> > name of the device it represents.
> > VDSM expects them to me in sync.
> >
> > Please provide the ifcfg files before and after you add the host to Engine.
> 
> Perhaps Christoph refers to the problem that [1] was meant to solve?
> 
> [1] https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/

To add on what didi says, this should be the default with el7's systemd.
It is surprising that your nics are named eth*, and not by the
predictable nic name scheme.

Maybe if you share your `lspci -vvv` and /var/log/messages of two
different boots, we can have a hint regarding your instability.



More information about the Users mailing list