[ovirt-users] Issue / strange behavior with GlusterFS nodes

[René Koch]

On 06/06/2014 11:21 AM, Sahina Bose wrote:
>
> On 06/06/2014 02:03 PM, Joop wrote:
>> René Koch wrote:
>>> Hi,
>>>
>>> I have the following oVirt setup:
>>> 2 storage hosts with GlusterFS (replicated) - will be extended to 4
>>> hosts (distributed-replicated)
>>> 4 virtualization hosts
>>> 1 datacenter
>>> 2 clusters (1 only for virtualization, 1 only for GlusterFS)
>>>
>>> The strange thing is now:
>>> One of the 2 GlusterFS hosts (these hosts and the Gluster volumes are
>>> managed by oVirt) fails to mount one of the storage domains, but I
>>> can mount it manually.
>>>
>>> So the first question is: why should these hosts mount the storage
>>> domains? They are in a cluster without virtualization, so it's not
>>> possible to run virtual machines on these hosts. I think they try to
>>> mount the storage domains because the storage domains are attached to
>>> the data center, but imho the cluster should be checked as well if
>>> virtualization is enabled.
>>>
>>> So I was thinking of moving the GlusterFS cluster to a new data
>>> center without storage domains, but this is impossible as Gluster
>>> volumes are created. The only possibility would be to remove the
>>> Gluster volume.
>>>
>>> So do you have any idea how I can solve this issue? I mean the
>>> Gluster volume is working fine, but it's really disturbing that I get
>>> an error message every 5 minutes.
>>>
>> Just to let you know you're not the only one with this problem. I have
>> had it too a couple of times and still don't know the right
>> incantation to get rid of it. I have the same setup: 2 clusters,
>> storage and virt, 2 nodes each, one datacenter. Currently on 3.4.1 and
>> had the problem on the 3.2/3 series. For me it seemed related to
>> upgrading but that could be just chance. It might as well be related
>> to a complete shutdown/restart of all machines ( this test setup is
>> not behind a ups which gives some rare opportunities to troubleshoot
>> problems :-) )
>
>
> Yes, this is a bug I would think. For a gluster only cluster, the
> storage domain is not relevant.
>
> Could you log a bug with the error attached, and we will try to fix
> this. If it's a simple enough fix, we could backport it to 3.4

Done: https://bugzilla.redhat.com/show_bug.cgi?id=1105513

I used component vdsm, but maybe it should belong to webadmin - I'm a 
bit unsure which component is responsible for this.


Thanks,
René

>
> thanks!
> sahina
>
>>
>> Joop
>>
>> _______________________________________________
>> Users mailing list
>> Users at ovirt.org
>> http://lists.ovirt.org/mailman/listinfo/users
>