Create snapshot in primary backend after done failover-host (Storwize)

[YuZhang]

Currently, after failover-host using Storwize backend, the host of the volume is not change, which is still the primary backend information. then create snapshot or clone a volume would lead those create in primary backend, which expect to create in replication_device backend.

Repro steps:
1. configure to support replication.
2. create a replication enable volume: cinder create --volume-type replication-enable-type 1
3. failover the host: cinder failover-host --backend_id backend-id host-name
4. create snapshot or clone a volume of one of the failed-over volume
5. check the place the snapshot or volume created, it is created on the primary backend(sv1 as example), which should create on the secondary backend(sv2).

cinder config like:
[sv1]
san_login = admin
san_password = admin
san_ip = 192.168.0.11
volume_driver = cinder.volume.drivers.ibm.storwize_svc.storwize_svc_iscsi.StorwizeSVCISCSIDriver
volume_backend_name = sv1
storwize_svc_volpool_name=cinder
replication_device = managed_backend_name:second_host@sv2#sv2,
                                       backend_id:svc_backend_id,
                                       replication_mode:global,
                                       san_ip:192.168.0.12,san_login:admin, san_password:admin,pool_name:cinder_target

[sv2]
san_login = admin
san_password = admin
san_ip = 192.168.0.12
volume_driver = cinder.volume.drivers.ibm.storwize_svc.storwize_svc_iscsi.StorwizeSVCISCSIDriver
volume_backend_name = sv2
storwize_svc_volpool_name=cinder_target

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/292570
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=693c91467afbcb92b29493ead7c8bfc455f41c81
Submitter: Jenkins
Branch: master

commit 693c91467afbcb92b29493ead7c8bfc455f41c81
Author: Vincent Hou <email address hidden>
Date: Thu Mar 17 14:11:43 2016 -0400

    Storwize: do the correct update after failover

    This patch adds the following changes:
    * If the service sets up during the fail-over, the storwize driver
    should take the corresponding secondary backend instead of the
    primary backend.

    * If the backend has failed over, the storwize
    should switch the primary helpers to the secondary.

    * Give the right parameter to delete the relationship between
    the primary and the secondary volumes.

    Change-Id: I7bf84bf6ebf4ecedcc75954b993e533c8ce404f8
    Closes-Bug: #1555945

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/cinder 10.0.0.0b3 development milestone.