[ovirt-users] Recommended gluster setup

Sahina Bose sabose at redhat.com
Fri Sep 15 12:30:25 UTC 2017


On Fri, Sep 15, 2017 at 5:48 PM, Alex K <rightkicktech at gmail.com> wrote:

> In the replica 3 + 1 arbiter does this mean that if I loose the two nodes
> having the normal volumes and left only with the node that has the arbiter
> volume, I loose all data?
>

Yes!


>
> Thanx,
> Alex
>
> On Fri, Sep 15, 2017 at 11:25 AM, Kasturi Narra <knarra at redhat.com> wrote:
>
>> Hi ,
>>
>>    yes, you are right. Since arbiter brick  has only metadata and  data
>> for the vm has to be served from one of the other two replicas, read is
>> slow.
>>
>>     Arbiter is a special subset of replica 3 volumes and is aimed at
>> preventing split-brains and providing same consistency as a normal replica
>> 3 volume with out consuming 3x space. You could use replica 3 and no issues
>> with that.
>>
>> Thanks
>> kasturi
>>
>> On Fri, Sep 15, 2017 at 12:41 PM, Abi Askushi <rightkicktech at gmail.com>
>> wrote:
>>
>>> Hi all,
>>>
>>> I see in the ovirt guides that a gluster volume replica 3 with 1 arbiter
>>> is recommended.
>>> Why not simple replica 3? Is it due to the higher replication data that
>>> would cause performance issues?
>>>
>>> What I am observing is that a VM running on the server which has the
>>> arbiter brick has slower read performance then when the same VM runs on
>>> another server with a normal brick. Has anyone observed this? Is it because
>>> the arbiter does not have the real data on it?
>>>
>>> Thanx,
>>> Alex
>>>
>>> _______________________________________________
>>> 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
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20170915/3baf7870/attachment.html>


More information about the Users mailing list