Issues with NetApp Cinder driver

[yocum137]

About 50% of the time, the NetApp Cinder driver (7-mode) does not create a custom parent volume to create iSCSI LUNs, instead using a volume created for NFS sharing of glance images.

For instance, the netapp cinder driver should create a volume based on the tenant UUID like this:

/vol/OpenStack_498561f42616446bbd405c1edb66f0cf

Then it should create a single volume/lun under that parent volume for a user created volume, like this:

volume-13ab14be-415a-48e5-9d1d-c079995b2a4a-BAAqB]CHFg4m

The resulting LUN looks like this:

/vol/OpenStack_498561f42616446bbd405c1edb66f0cf/volume-13ab14be-415a-48e5-9d1d-c079995b2a4a-BAAqB]CHFg4m

This happens about 50% of the time.  The other 50% the driver defaults to a volume currently used for glance images (/vol/vol2):

/vol/vol2/volume-77bd0ea6-2adc-4dd5-80a8-8dce8abecb27-BAAqB]CHFg4H

This became broken in the last few months - I can't pinpoint exactly when.  Here are the current RPMs installed on the cinder-volume (controller) node:

openstack-cinder-2014.1.4-1.1.el6ost.noarch
python-cinderclient-1.0.9-1.el6ost.noarch
python-cinder-2014.1.4-1.1.el6ost.noarch

I need this a) fixed and b) a way to migrate iscci volumes out of the glance NFS volume into tenant specific volumes.

Dan

dyocum@redhat.com

[rcallawa]

Our Cinder drivers don't create new FlexVol volumes, and your description indicates that perhaps you don't have netapp_volume_list set correctly in your cinder.conf file? This setting would limit the scope of provisioning for Cinder into the names of the FlexVol volumes you list in the config option (http://netapp.github.io/openstack-deploy-ops-guide/icehouse/content/cinder.7mode.iscsi.configuration.html)

The behavior didn't change dramatically between Havana and Icehouse (2014.1) so I am curious as to what version of OpenStack you were running before.

As to volume migration, this page describes how you can move the LUN from 1 FlexVol to another - https://kb.netapp.com/support/index?id=1011238&page=content&locale=en_US

[yocum137]

Hey rcallawa, sorry for the delay in replying - I'm not sure why I didn't get an email when you replied...

Anyway,  in reply to your questions, this cloud started life on Folsom and has been upgraded through Grizzly -> Havana -> Icehouse on RHEL6.

netapp_volume_list is unset in cinder.conf.  😞

I suggest that if netapp_volume_list is unset, then NO volume should be used, that 'cinder create' should fail until it is set.

In Folsom/Grizzly era the netapp driver would create (at least) one volume per tenant.  I assumed this was to enforce tenant quotas.  How are tenant quotas enforced now that we're supposed to specify the FlexVol all the iscsi LUNs are supposed to get created in?

Ugh - 549 LUNs to move out of the /var/lib/glance NFS volume.  

This is going to hurt.  😞