On Thu, Mar 17, 2016 at 10:41 AM, Nicolas Ecarnot <nicolas(a)ecarnot.net> wrote:
Le 17/03/2016 09:22, Edward Haas a écrit :
>
> On Thu, Mar 17, 2016 at 9:23 AM, Nicolas Ecarnot <nicolas(a)ecarnot.net>
> wrote:
>>
>> Le 17/03/2016 07:32, Edward Haas a écrit :
>>>
>>>
>>> Hello Nicolas,
>>>
>>> Please let us know which docs you refer to.
>>
>>
>>
>>
>>
http://community.redhat.com/blog/2014/11/up-and-running-with-ovirt-3-5-pa...
>>
>>> Engine by itself should work fine with NM.
>>
>>
>>
>> I'm installing my 5th 3.6.3 oVirt DC this month, and I'm still
witnessing
>> that when keeping NM_CONTROLLED=yes in my
>> /etc/sysconfig/network-script/ifcfg-blahblahblah files, the network setup
>> doesn't resist a reboot (we're using bonding + vlan, and it all goes
well
>> with NO networkmanager).
>
>
> Do you refer to the Host/s (which run the hypervisor with the VM/s) or
> the Engine?
I'm witnessing these issues on the hosts.
I also tried to use nmtui, but no method made it until I switched back to
using vim and disable nm in the ifcfg-* files.
Then, oVirt made a good work understanding, managing, and syncing these
files (bond + vlan).
No big issue then, but I'd like to be sure. I'm about to add a cf-engine
promise to disable NetworkManager on the hosts.
--
Nicolas ECARNOT
On the hosts we do not support NM at the moment. When you add a host
through Engine, it is supposed to disable NM on the host.
The problem with NM on the host is that it can take ownership of
devices from VDSM in specific scenarios.
There are plans to address this issue and allow NM to run in parallel,
but that is for the long run.
There should not be such a limitation on the Engine side.
Thanks,
Edy.