[Openstack] Why GlusterFS should not be integrated with OpenStack

John Mark Walker johnmark at johnmark.org
Tue Sep 10 14:59:16 UTC 2013

On Tue, Sep 10, 2013 at 10:42 AM, Caitlin Bestler <
caitlin.bestler at nexenta.com> wrote:

>  I'm an architect at Nexenta. So not surprisingly I fully agree that
> Gluster is a poor choice for storage in OpenStack.

Nexenta could easily apply GlusterFS as an aggregation/scale-out layer  -
in fact, one could argue they should. Your customers would benefit
enormously, so no, I don't understand your statement at all.

> However the real question is what criteria storage vendors should be
> judged by the project. These should be
> clearly stated and easily verified. Once compliance is dealt with, judging
> the quality of specific solutions and
> their suitability to specific needs is something that the end users should
> do, not the openstack project.


> The Cinder project has set minimum standards for Volume Drivers for the
> Havana release. Those standards should be applied without debating their
> specifics for Gluster on the mailing list.

Are you suggesting that all storage backends are the same? Because that's
not true. Data without silos is unique to the GlusterFS implementation. I'm
not sure how I can discuss GlusterFS as a storage solution without also
mentioniong that feature.

> I would be in full agreement, however, that OpenStack should not
> accommodate an API tailored to Gluster's unique architecture.

No one has mentioned that. We're talking about separating implementation
from the API interface.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20130910/c229eb3f/attachment.html>

More information about the Openstack mailing list