[Openstack] Openstack] Suggestions for shared-storage cluster file system

Greg Chavez greg.chavez at gmail.com
Wed Feb 20 01:17:23 UTC 2013


So it sounds like what we're talking about here is running a Ceph or
GlusterFS node alongside the compute node (ie, on the same physical
system).  I assume then that VMs access their volumes via NFS or iSCSI from
Cinder on the controller node, and in turn, Cinder reads and writes to the
cluster FS, which is spread out among the physical data volumes on the
compute nodes.  Is this correct?


On Fri, Feb 15, 2013 at 2:29 PM, Sébastien Han <han.sebastien at gmail.com>wrote:

> Hi,
>
>
> Important: Mount the CephFS filesystem on the client machine, not the
>> cluster machine.
>
>
> It's just like NFS, if you mount an NFS export on the NFS server, you get
> kernel locks.
>
> Unfortunately even if love Ceph far more than the other, I won't go with
> CephFS, at least not know. But if are in a hurry and looking for a DFS then
> GlusterFS seems to be a good candidate. NFS works pretty well too.
>
> Cheers.
>
> --
> Regards,
> Sébastien Han.
>
>
> On Fri, Feb 15, 2013 at 4:49 PM, JuanFra Rodriguez Cardoso <
> juanfra.rodriguez.cardoso at gmail.com> wrote:
>
>> Another one:
>>
>>  - MooseFS (
>> http://docs.openstack.org/trunk/openstack-compute/admin/content/installing-moosefs-as-backend.html
>> )
>>  - GlusterFS
>>  - Ceph
>>  - Lustre
>>
>> Regards,
>> JuanFra
>>
>>
>> 2013/2/15 Samuel Winchenbach <swinchen at gmail.com>
>>
>>>  Hi All,
>>>
>>> Can anyone give me a recommendation for a good shared-storage cluster
>>> filesystem?   I am running kvm-libvirt and would like to enable live
>>> migration.
>>>
>>> I have a number of hosts (up to 16) each with 2xTB drives.  These hosts
>>> are also my compute/network/controller nodes.
>>>
>>> The three I am considering are:
>>>
>>> GlusterFS - I have the most experience with this, and it seems the
>>> easiest.
>>>
>>> CephFS/RADOS - Interesting because glance supports the rbd backend.
>>>  Slightly worried because of this though "Important:
>>> Mount the CephFS filesystem on the client machine, not the cluster
>>> machine."
>>> (I wish it said why...)  and "CephFS is not quite as stable as the block
>>> device and the object storage gateway."
>>>  Lustre - A little hesitant now that Oracle is involved with it.
>>>
>>>
>>> If anyone has any advice, or can point out another that I should
>>> consider it would be greatly appreciated.
>>>
>>> Thanks!
>>>
>>> Sam
>>>
>>>
>>> _______________________________________________
>>> Mailing list: https://launchpad.net/~openstack
>>> Post to     : openstack at lists.launchpad.net
>>> Unsubscribe : https://launchpad.net/~openstack
>>> More help   : https://help.launchpad.net/ListHelp
>>>
>>>
>>
>> _______________________________________________
>> Mailing list: https://launchpad.net/~openstack
>> Post to     : openstack at lists.launchpad.net
>> Unsubscribe : https://launchpad.net/~openstack
>> More help   : https://help.launchpad.net/ListHelp
>>
>>
>
> _______________________________________________
> Mailing list: https://launchpad.net/~openstack
> Post to     : openstack at lists.launchpad.net
> Unsubscribe : https://launchpad.net/~openstack
> More help   : https://help.launchpad.net/ListHelp
>
>


-- 
\*..+.-
--Greg Chavez
+//..;};



-- 
\*..+.-
--Greg Chavez
+//..;};
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20130219/a9fc7cd1/attachment.html>


More information about the Openstack mailing list