[Users] Network not sync and problems reconfiguring

[Lior Vernia]

On 10/01/14 16:16, Juan Pablo Lorier wrote:
> 
> On 10/01/14 11:59, Lior Vernia wrote:
>>
>> On 10/01/14 15:54, Juan Pablo Lorier wrote:
>>> On 10/01/14 11:13, Lior Vernia wrote:
>>>> On 10/01/14 11:53, Juan Pablo Lorier wrote:
>>>>> Hi Lior,
>>>>>
>>>>> I'll try to elaborate between lines:
>>>>>
>>>>> On 09/01/14 18:31, Lior Vernia wrote:
>>>>>> Hi Juan,
>>>>>>
>>>>>> I'm not sure I understand a couple of things, could you elaborate?
>>>>>>
>>>>>> 0. What version engine are you running?
>>>>> 3.3.1 when the problem appeared, but now I moved to 3.3.2
>>>>>> 1. "Out of the blue" - no trigger that you can think of, that would help
>>>>>> retrace the cause?
>>>>> The only thing I can think of was a reboot (clean) of the engine VM (not
>>>>> self hosted, just a vbox VM). After the problem appeared, I started
>>>>> making many changes and tests so I can't recreate the exact scenario at
>>>>> this time.
>>>> Okay, currently nothing springs to mind as to what may have caused this,
>>>> I'll try to think on it some more.
>>>>
>>>>>> 2. From your other e-mail, I understand that somehow the management
>>>>>> network lost its VLAN tagging, and that it shows as out-of-sync in the
>>>>>> Setup Host Networks dialog. Correct?
>>>>> Yes. Also other LN where out of sync, but the hosts had differences in
>>>>> which LN went out of sync. All of them have ovirtmgmt out of sync.
>>>>>> 3. Do I understand correctly that even though it's marked as
>>>>>> out-of-sync, you don't see the "synchronize" checkbox when editing the
>>>>>> network on the host (little pencil icon)?
>>>>> It doesn't. I made some tests and I found that it's a problem with
>>>>> firefox (mine is version 26 over Fedora 19). It's showing things a
>>>>> little bit bigger than chrome and so the sync option is not shown in the
>>>>> edit window. Other functions also are not working properly, for
>>>>> instance, if I choose a particular item into a category (host, vm, etc)
>>>>> in the left tree, the right frame does not change, I have to choose the
>>>>> group (hosts, vms, etc) and choose the particular item among all the
>>>>> listed in the right frame. I've checked and firefox 22 on windows has
>>>>> the same behavior. This started to happen after I upgraded from 3.3.0 to
>>>>> 3.3.1 but never saw the lack of the sync option until now.
>>>> My machine currently isn't in a state to test these things on different
>>>> versions of Firefox. If you can pin-point exactly on what versions this
>>>> occurs, I would greatly appreciate it if you could open a bug about it
>>>> in bugzilla.redhat.com and I'll try to get it fixed.
>>>>
>>>> It does sound that it's only a graphical issue that's causing the
>>>> checkbox to be hidden. So as an immediate workaround, you could use
>>>> either the Firefox web developer tools or the Firebug plugin to enlarge
>>>> the dialog and panels enough so that the checkbox shows, then mark it
>>>> and have the networks synchronized. Please let me know if that works for
>>>> you.
>>> I'll open a bug then. The versions I tested were on the last mail, I'll
>>> put that in the bug also.
>>> My work arround was simpler, just did it with chrome :-)
>>> Anyway, this fix only the sync stuff, not the need of modifying the
>>> ovirtmgmt to be again a tagged network. Is there a hack to the db or
>>> something like that I can do to set this back without destroing the hole
>>> DC (which I can't do either because of a dead lock with the flow of
>>> dependencies in ovirt and opened a BZ for that).
>>> Regards,
>>>
>> Can't you edit the management network to be VLAN-tagged, and then when
>> it appears as out-of-sync on the host mark it to be synchronized?
>>
> Webadmin won't let me as it's been used in hosts (tried). It's not
> allowed until 3.4 as I understand.
> 

Ahhh yes, I recall the problem now (thought it might be possible as long
as it isn't attached to VMs). You could indeed hack the DB. Setting the
proper VLAN tag for ovirtmgmt in the "network" table should suffice
(assuming the VLAN interfaces are still intact on the hosts, as I deduce
from ovirtmgmt being marked as out-of-sync).

> 
>>>>>> Yours, Lior.
>>>>>>
>>>>>> On 09/01/14 17:44, Juan Pablo Lorier wrote:
>>>>>>> Hi,
>>>>>>>
>>>>>>> Out of the blue some of the hosts got some LNs out of sync (not the same
>>>>>>> in all of them). If I try to resync the LN, it does not show a resync
>>>>>>> option, so I took one host and detached the lns and reconfigured them.
>>>>>>> After that, ovirtmgmt is not created correctly, it does not create the
>>>>>>> vlan interface on top of the bond and also, it's adding the bond to the
>>>>>>> bridge instead of the vlan interface.
>>>>>>> I got it partially running but it's still non operational as the engine
>>>>>>> complains the host can't mount the gluster domain (it can reach the
>>>>>>> network of the other brick):
>>>>>>>
>>>>>>> [root at ovirt4 network-scripts]# gluster volume status
>>>>>>> Status of volume: glusterfs
>>>>>>> Gluster process                        Port    Online    Pid
>>>>>>> ------------------------------------------------------------------------------
>>>>>>> Brick 192.168.128.82:/glusterfs                49154    Y    13288
>>>>>>> Brick 192.168.128.83:/glusterfs                49154    Y    5961
>>>>>>> NFS Server on localhost                    2049    Y    4557
>>>>>>> Self-heal Daemon on localhost                N/A    Y    4558
>>>>>>> NFS Server on 192.168.128.82                2049    Y    15815
>>>>>>> Self-heal Daemon on 192.168.128.82            N/A    Y    15819
>>>>>>>
>>>>>>>
>>>>>>> I'm not synchronizing any other host just in case this happens to the
>>>>>>> other host too.
>>>>>>> Regards,
>>>>>>> _______________________________________________
>>>>>>> Users mailing list
>>>>>>> Users at ovirt.org
>>>>>>> http://lists.ovirt.org/mailman/listinfo/users
>>>>>>>
>