violin v6000 drivers query wrong backend location during export verify

[Ryan Lucio]

After performing a lun or snapshot export, the violin drivers enter a short verification loop to wait for the lun to become available on the wire. As it turns out, the backend state information being checked is not appropriate for this use and can actually cause the verification loop to return prematurely thus causing nova connections to the lun/snap to fail.

The appropriate place to check is actually the backend state that reflects the export serial number generated by the kernel. This serial number (usn_id) is not generated until the lun is fully exported.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/165655

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/165655
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=ad5eb92dcd5afcbce99c5c4812415fdcb6d8b339
Submitter: Jenkins
Branch: master

commit ad5eb92dcd5afcbce99c5c4812415fdcb6d8b339
Author: Ryan Lucio <email address hidden>
Date: Thu Mar 12 16:12:10 2015 -0700

    VMEM v6000: Fix export verify routines

    New 'state' nodes in vmos6 are reflecting the kernel/system state
    indicating lun accessibility on the data plane. The nodes used
    previously were only reflecting success on the mgmt plane, meaning
    that the routines were sometimes returning before the target/export
    was actually available for client use.

    Also, switch to using one global iscsi target group (backend
    container that groups target IPs) to avoid a backend export bug
    exposed by the changes mentioned above.

    Change-Id: I1cd296d6d02562e1eb7374df736b2649fe2692a2
    Closes-Bug: #1430014