[openstack-dev] The State of the NFS Driver ...

Matt Riedemann mriedem at linux.vnet.ibm.com
Wed Aug 31 01:50:26 UTC 2016

On 8/30/2016 10:50 AM, Jay S. Bryant wrote:
> All,
> I wanted to follow up on the e-mail thread [1] on Cloning support in the
> NFS driver.  The purpose of this e-mail is to provide the plan for the
> NFS driver going forward as I see it.
> First, I am aware that the driver has gone quite some time without care
> and feeding.  For a number of reasons, the Public Cloud team within IBM
> is currently dependent upon the NFS driver working properly for the
> cloud environment we are building.  Given our current dependence on the
> driver we are planning on picking up the driver and maintaining it.
> The first step in this process was getting the existing patch that adds
> snapshot support for NFS [2] rebased.  I did this work a couple of weeks
> ago and also got all the unit tests working for the unit test
> environment on the master branch.  I now see that it is in merge
> conflict again, I plan to continue to keep the patch up-to-date.
> Erlon has been investigating issues with attaching snapshots.  It
> appears that this may be related to AppArmor running on the system where
> the VM is running and attachment is being attempted.  I am hoping to
> look into the other questions posed in the patch review in the next week
> or two.
> The next step is to create a dependent patch, upon the snapshot patch,
> to implement cloning.  I am planning to also undertake this work.  I am
> assuming that getting the cloning support in place shouldn't be too
> difficult once snapshots are working as it will be just a matter of
> using the support from the remotefs driver.
> The last piece of work we have in flight is working on adding QoS
> support to the NFS driver.  We have the following spec proposed to get
> that work started: [3]
> So, we are in the process of bringing the NFS driver up to good
> standing.  During this process we would greatly appreciate reviews and
> input from those of you who have previously worked on the driver in
> order to expedite integration of the necessary changes. I feel it is in
> the best interest of the community to get the driver updated and
> supported given that it is the 4th most used driver according to our
> user survey.  I think it would not look good to our users if it were to
> suddenly be removed.
> Thanks to all of your for your support in this effort!
> Jay
> [1]
> http://lists.openstack.org/pipermail/openstack-dev/2016-August/102193.html
> [2] https://review.openstack.org/#/c/147186/
> [3] https://review.openstack.org/361456
> __________________________________________________________________________
> 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

IMO priority #1 is getting the NFS job passing consistently, who is 
working on that? Last I checked it was failing a bunch because it was 
running snapshot and clone tests, which obviously don't work since that 
support isn't implemented in the driver. I think configuring tempest in 
the devstack-plugin-nfs repo is fairly straightforward, someone just 
needs to do it.

But at least that gets you closer to a clean NFS job run which gets it 
out of the experimental queue (possibly) and as a non-voting job in 
Cinder so you can see if you're regressing anything (or if anything else 
regresses it once you have clean CI runs).

My 2 cents.



Matt Riedemann

More information about the OpenStack-dev mailing list