[openstack-dev] [Cinder] LIO support in Cinder
Russell Bryant
rbryant at redhat.com
Wed Nov 28 14:54:14 UTC 2012
On 11/27/2012 11:12 PM, John Griffith wrote:
>
>
> On Tue, Nov 27, 2012 at 12:04 PM, Eric Harney <eharney at redhat.com
> <mailto:eharney at redhat.com>> wrote:
>
> On Mon, 2012-11-26 at 20:22 -0500, Russell Bryant wrote:
> > On 11/26/2012 06:46 PM, Eric Harney wrote:
> > > Hi everyone,
> > >
> > > I've posted a blueprint for supporting the LIO (linux-iscsi.org
> <http://linux-iscsi.org>) target
> > > in Cinder. This leverages targetd to interface with LIO, and
> will be
> > > implemented as an alternative target within the iSCSI driver.
> > >
> > > A detailed spec has not been posted yet, but I would like to welcome
> > > high-level feedback on this as a feature and start discussion
> around it.
> > >
> > >
> https://blueprints.launchpad.net/cinder/+spec/lio-support-via-targetd
> >
> > Thanks, Eric!
> >
> > I have one implementation question. You mention that this will be
> > implemented as an alternative target within the iSCSI driver. Why not
> > make it its own driver?
> >
> > IIRC, the iSCSI driver is built around exposing local storage. A
> > targetd driver would be managing remote storage resources, so it seems
> > to be a different model.
> >
> > Thoughts?
> >
>
> Thinking about this a bit more, it does probably make sense for it to be
> its own driver, which would be consistent with many of the other drivers
> for remote storage resources.
>
> I'm not too clear on where the line is here (between iSCSI vs. other
> driver), but your reasoning sounds right to me.
>
>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> <mailto:OpenStack-dev at lists.openstack.org>
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
>> I'm curious as to the rationale of including a (relatively new) python project
>> for this. It seems like the fastest path to using LIO would be to simply
>> extend the tgt code to be able to use targetcli[1] in addition to tgtadmin
>> and ietadmin. That seems like literally a few days of work.
>
> I have to say this (Vish's comment above) is actually more of what I had
> envisioned when I talked with Eric about this last week. I also think
> that doing an iterative approach is a good idea.
>
> As far as targetd, I'm not sure it's quite ready (well definitely don't
> think it's ready in the Grizzly time-frame)? If I'm wrong on this
> please let me know, but in terms of distro support aren't we looking at
> 2013 releases before we start seeing it fully integrated?
Andy Grover responded about the status of targetd, but I guess it's
stuck in list moderation. Quoted below.
The first Fedora release it will be in is Fedora 18, and then it will be
in RHEL-7 later on (so yes, 2013).
A targetcli version in the short term seems ok, at least for testing/PoC
setups. It just may not be very pretty since it doesn't provide a good
way for error detection.
On 11/26/2012 10:16 PM, Andy Grover wrote:> On 11/26/2012 05:17 PM,
Russell Bryant wrote:
>> On 11/26/2012 07:25 PM, Huang Zhiteng wrote:
>>> This looks great! Supporting new target is definitely helpful to
>>> Cinder. I can't wait to try it out. And I'm wondering if you could
>>> introduce a little bit more on LIO, like what's advantage compared to
>>> TGT, IET.
>
> I'd say better performance, scalability, and multi-protocol support. See
> http://linux-iscsi.org/ for more details.
>
>>> One dumb question regarding to 'targetd' though, in its readme, it
>>> claimed itself as 'pre-alpha', is it ready to use?
>>
>> We got turned on to targetd by its author. I CC'd him on this message.
>> Hopefully he can comment further on the development status of targetd.
>
> Author here :)
>
> It's a new project, but I believe it is ready for widespread use. I've
> updated the targetd documentation to remove the 'pre-alpha' designation.
> targetd is in Fedora 18 and is planned for RHEL 7.
>
> Regards -- Andy
>
--
Russell Bryant
More information about the OpenStack-dev
mailing list