Open Stack

Ben,

Thank you very much for the history on the driver here.  This was all 
news to me.  It helps me to understand why cloning was listed as 
supported but was no longer implemented.  Also explains why the simpler 
'cp' based solution hasn't been implemented.  I was wondering why that 
hadn't just been done.

IBM is planning to take over the process of developing/maintaining the 
NFS driver, at least in the near term. Erlon has also been helping with 
this.

We would really like to avoid the driver being removed.  Perhaps I will 
send a separate 'State of the NFS driver' update e-mail to the mailing 
list to explain what I think the state of the driver is and what I see 
to be the plan going forward.

Thanks,

Jay


On 08/25/2016 12:06 PM, Ben Swartzlander wrote:
>
> Originally the NFS driver did support snapshots, but it was 
> implemented by just 'cp'ing the file containing the raw bits. This 
> works fine (if inefficiently) for unattached volumes, but if you do 
> this on an attached volume the snapshot won't be crash consistent at all.
>
> It was decided that we could do better for attached volumes by 
> switching to qcow2 and relying on nova to perform the snapshots. Based 
> on this, the bad snapshot implementation was removed.
>
> However, for a variety of reasons the nova-assisted snapshot 
> implementation has remained unmerged for 2+ years and the NFS driver 
> has been an exception to the rules for that whole time.
>
> I would like to see that exception end in the near future with either 
> the removal of the driver or the completion of the Nova-assisted 
> snapshot implementation, and it doesn't really matter to me which.
>
> There is a 3rd alternative which would be to modify the NFS driver to 
> require a specific filesystem that supports snapshots (there are a few 
> choices here, but definitely NOT ext4). Unfortunately those of us who 
> work for storage vendors aren't motivated to make such a modification 
> because it would be effectively creating more competition for 
> ourselves. The only way this could happen is if someone not working 
> for a storage vendor takes this on.
>
> -Ben
>
> On August 25, 2016 10:39:35 AM Erlon Cruz <sombrafam at gmail.com> wrote:
>
>> Hi Jordan, Slade,
>>
>> Currently NFS driver does not support cloning neither snapshots 
>> (which are the base for implementing cloning). AFAIC, the NFS driver 
>> was in Cinder before the minimum requirements being discussed and 
>> set, so, it just stood there with the features it already supported.
>>
>> There is currently this job 
>> 'gate-tempest-dsvm-full-devstack-plugin-nfs-nv'[1] that by the way 
>> are failing in the same test you mentioned tough passing the snapshot 
>> tests (not shure how the configuration is doing that) and a work[2] 
>> in progress to support the snapshot feature.
>>
>> So, Jordan, I think its OK to allow tempest to skip this tests, 
>> provided that at least in the NFS driver, tempest isn't being an 
>> enforcement to Cinder minimum features requirements.
>>
>> Erlon
>>
>>
>> [1] 
>> http://logs.openstack.org/86/147186/25/experimental/gate-tempest-dsvm-full-devstack-plugin-nfs-nv/b149960/
>> [2] https://review.openstack.org/#/c/147186/
>>
>> On Wed, Aug 24, 2016 at 6:34 PM, Jordan Pittier 
>> <jordan.pittier at scality.com <mailto:jordan.pittier at scality.com>> wrote:
>>
>>
>>     On Wed, Aug 24, 2016 at 6:06 PM, Slade Baumann
>>     <baumann at us.ibm.com <mailto:baumann at us.ibm.com>> wrote:
>>
>>         I am attempting to disable clone tests in tempest as they aren't
>>         functioning in NFS. But the tests test_volumes_clone.py and
>>         test_volumes_clone_negative.py don't have the "clone" feature
>>         toggle in them. I thought it obvious that if clone is disabled
>>         in tempest, the tests that simply clone should be disabled.
>>
>>         So I put up a bug and fix for it, but have been talking with
>>         Jordan Pittier and he suggested I come to the mailing list to
>>         get this figured out.
>>
>>         I'm not asking for reviews, unless you want to give them.
>>         I'm simply asking if this is the right way to go about this
>>         or if there is something else I need to do to get this into
>>         Tempest.
>>
>>         Here are the bug and fix:
>>         https://bugs.launchpad.net/tempest/+bug/1615770
>>         <https://bugs.launchpad.net/tempest/+bug/1615770>
>>         https://review.openstack.org/#/c/358813/
>>         <https://review.openstack.org/#/c/358813/>
>>
>>         I would appreciate any suggestion or direction in this problem.
>>
>>         For extra reference, the clone toggle flag was added here:
>>         https://bugs.launchpad.net/tempest/+bug/1488274
>>         <https://bugs.launchpad.net/tempest/+bug/1488274>
>>
>>     Hi,
>>     Thanks for starting this thread. My point about this patch is, as
>>     "volume clone" is part of the core requirements [1] every Cinder
>>     drive must support, I don't see a need for a feature flag. The
>>     feature flag already exists, but that doesn't mean we should
>>     encourage its usage.
>>
>>     Now, if this really helps the NFS driver (although I don"t know
>>     why we couldn't support clone with NFS)... I don't have a strong
>>     opinion on this patch.
>>
>>     I -1ed the patch for consistency: I agree that there should be a
>>     minimum set of features expected from a Cinder driver.
>>
>>     [1]
>>     http://docs.openstack.org/developer/cinder/devref/drivers.html#core-functionality
>>     <http://docs.openstack.org/developer/cinder/devref/drivers.html#core-functionality>
>>
>>     Cheers,
>>     Jordan
>>
>>     <http://bit.ly/2aKbaTu>
>>     __________________________________________________________________________
>>     OpenStack Development Mailing List (not for usage questions)
>>     Unsubscribe:
>>     OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
>>     <http://OpenStack-dev-request@lists.openstack.org?subject:unsubscribe>
>>     http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>     <http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev>
>>
>>
>> __________________________________________________________________________
>> OpenStack Development Mailing List (not for usage questions)
>> Unsubscribe: OpenStack-dev-request at lists.openstack.org 
>> <mailto:OpenStack-dev-request%40lists.openstack.org>?subject:unsubscribe
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20160830/4f32cc3d/attachment.html>