[Users] Anyone started working on the documentation for GlusterFS and Ovirt 3.1 yet?

Shireesh Anjal sanjal at redhat.com
Mon Jun 11 06:11:54 UTC 2012


On Sunday 10 June 2012 02:02 AM, Robert Middleswarth wrote:
> On 06/09/2012 11:33 AM, Haim Ateya wrote:
>> adding Shireesh\Daniel
>>
>> Hi Robert,
>>
>> afaik, there is no official documentation that show's how to work
>> with glusterFS using oVirt since its too new and we are in early 
>> integration phases (shireesh, correct me if i'm wrong).
> That is why I asked is anyone has started working on the doc's

Here is the feature page: http://ovirt.org/wiki/Features/Gluster_Support
We intend to add more documentation soon, though nothing else is 
available right now.

>> I would start by browsing tohttp://www.gluster.org/  and read about 
>> the underlying technology (glusterFS, admin guide is good place to 
>> start).
> Read though there doc's but wasn't sure how much needed to be done 
> using native glusterfs tools and how much is done using ovirt engine 
> because of the fact it was recently integrated means there are still 
> likely holes in the process.

Most of the features mentioned in the feature page, except async tasks 
(rebalance volume, migrate brick) and bootstrapping changes are 
currently available from UI and REST api. You should run engine with the 
option "InstallVds" set to false and SSL disabled (on both engine and 
vdsm). When creating clusters, make sure you select the checkbox to 
enable gluster services, and always add hosts to such clusters from UI only.

>> as for tips and hints, I will describe what you need in general (very 
>> high-level):
>>
>> - cluster and data-center - edit cluster and set the gluster check box
>> - host side (vds):
>>    * install vdsm-gluster package
>>    * install all gluster-related packages
>>    * make sure all gluster related services are fully operational
> Didn't see vdsm-gluster when I built the nodes but when I looked for 
> the package it was there.  Already had all the glusterfs services 
> running.
>> - create new volume (volume is a collection of bricks which represent 
>> a collection of export domains):
>>    * allocate storage space on your host
>>    * make sure this space is exposed via export (NFS) directory
>>    * create new volume\bricks
> This is where I am confused.  glusterfs supports 3 methods of exposing 
> the cluster.  Is ovirt using the NFS or the native(fs) method?  If it 
> is NFS why the interface for bricks?  Wouldn't you just create a giant 
> nfs share?  This is the area I am most confused with and looking for 
> hints on.

To be more precise, glusterfs supports 3 methods of exposing a *volume*, 
and not cluster. A cluster can contain a number of volumes. In Gluster 
terminology, Bricks (server:directory) are building blocks of a 
"volume". Volumes can be exported using native (GLUSTER), NFS or CIFS 
protocols, though only GLUSTER and NFS are available from the UI for 
now. A volume need not necessarily have bricks on all servers of the 
cluster. If you want one giant NFS share spread over all servers of the 
cluster, you'll have to use one brick from each of the servers when 
creating the volume, and then export the Volume via NFS.

Feel free to ask glusterfs specific questions on the #gluster IRC 
channel on FreeNode.

>
> Thanks
> Robert
>> Thanks,
>>
>> Haim
>> ----- Original Message -----
>>> From: "Robert Middleswarth"<robert at middleswarth.net>
>>> To:users at ovirt.org
>>> Sent: Friday, June 8, 2012 9:23:40 PM
>>> Subject: [Users] Anyone started working on the documentation for 
>>> GlusterFS and Ovirt 3.1 yet?
>>>
>>> I installed ovirt 3.1 using the assume repo provided by
>>> http://www.dreyou.org/ovirt/  and it works really well.  Been a few
>>> glitches but overall it works well.  I see an interface for GlusterFS
>>> is
>>> in there but no clue how to use it any hints or info been created
>>> yet?
>>>
>>> Thanks
>>> Robert
>>>
>>> _______________________________________________
>>> 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/20120611/e3723b2b/attachment-0001.html>


More information about the Users mailing list