[ovirt-users] Is it a plausible configuration?

[Kiril L]

This should clear things up a bit -
https://joejulian.name/blog/glusterfs-split-brain-recovery-made-easy/
even

On Thu, Jun 4, 2015 at 9:40 AM, Юрий Полторацкий
<y.poltoratskiy at gmail.com> wrote:
> Hi,
>
> Pardon me for jumping in subj, but...
>
> Could you explain me in a few words or give a link where to read about
> "split-brain"?
>
> I made a lab: cluster (virt and gluster services both) based on two server.
> With this options:
>  cluster.server-quorum-type none
>  cluster.quorum-type fixed
>  cluster.quorum-count 1
> I get a cluster that worked just fine even with only one host alive.
>
> I have run VM on a not SPM host B, have made some files, then on the SPM
> host A have stoped network service (host has gone, no power management),
> have made some huge files on the VM again, after that I have rebooted
> manualy the host A and there was no unexpected result. VM have worked
> without problem, I have rebooted the VM several times and there were no
> problems with any files I have made before.
>
> I want to build a cluster with only two hosts, because I have only two good
> servers free, and only 8x3TB HDD (for RAID10 in each host). To get one more
> server and 4HDD is unbelievable in a close future..
>
> Thanks.
>
>
> 2015-06-03 16:54 GMT+03:00 Simone Tiraboschi <stirabos at redhat.com>:
>>
>>
>>
>> ----- Original Message -----
>> > From: "Kiril L" <neohidra at gmail.com>
>> > To: users at ovirt.org
>> > Sent: Wednesday, June 3, 2015 3:32:52 PM
>> > Subject: [ovirt-users] Is it a plausible configuration?
>> >
>> > Would you please tell me if this configuration is doable or there is
>> > something that i am missing?
>> >
>> > I would like to use only two servers (X and Y) for VDI and Gluster
>> > based storage.
>> > Hosted engine for oVirt and replicated volumes between X and Y for the
>> > gluster storage. Is a third machine Z a must?
>>
>> It works also with just two hosts but it's not that safe: a replica-2
>> GlusterFS volume affected by a split-brain issue could be not self-healing
>> while with replica-3 you could rely on quorum enforcement just cause you are
>> using an odd host number.
>>
>> For oVirt 3.6 we are working on
>> http://www.ovirt.org/Features/Self_Hosted_Engine_Hyper_Converged_Gluster_Support
>>
>>
>> > _______________________________________________
>> > Users mailing list
>> > Users at ovirt.org
>> > http://lists.ovirt.org/mailman/listinfo/users
>> >
>> _______________________________________________
>> Users mailing list
>> Users at ovirt.org
>> http://lists.ovirt.org/mailman/listinfo/users
>
>