EMC VMAX terminate_connection unmaps volume from wrong VM

[Carl Pecinovsky]

The terminate_connection() processing contains this flow in driver:

_unmap_lun() -->_remove_members() -->remove_and_reset_members() -->get_associated_masking_group_from_device()

However, this fails to take into account the particular masking view that is associated with the volume's connection. In the PowerVC live migration use case, a volume can be mapped to source and target VMs (NPIV) or hosts for a window of time, and if so, get_associated_masking_group_from_device() can look up the wrong masking storage group since the view is not passed in. It likely should be passed down from its initial lookup in find_device_number(), and that method's lookup of the view needs to take into account the initiator wwpns per https://bugs.launchpad.net/cinder/+bug/1382641

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/140909

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/140909
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=d94cc2888c5097306c43de3b2b424b408d52161a
Submitter: Jenkins
Branch: master

commit d94cc2888c5097306c43de3b2b424b408d52161a
Author: Xing Yang <email address hidden>
Date: Mon Dec 15 11:47:44 2014 -0500

    Fixed a problem in terminate_connection in VMAX driver

    The VMAX driver unmaps the volume from the wrong VM in
    terminate_connection during live migration. This patche
    fixed this problem.

    Closes-Bug: #1395845
    Change-Id: Ida7136165b959a9e0d7d50c240b447bf237caa6c

[Carl Pecinovsky (csky)]

Xing,
The fix done for this defect appears to be for the initialize_connection flow. However, this defect was written for terminate_connection(). Can you please explain, since it seems the method mentioned in the description still needs fixing?

[Xing Yang (xing-yang)]

Carl,

There's actually a second part of this bug that is more involved. It is submitted here:

https://review.openstack.org/#/c/143891/

I marked it as WIP because it depends on two other fixes which are still being reviewed:

https://review.openstack.org/#/c/140913/
https://review.openstack.org/#/c/135165/

I should have said "partially fixes bug" instead of "Closes-Bug" when submitting the fix that was already merged.