failback causes cinder to enable backend which may be frozen before failover

[jiamin]

Steps to Reproduce:

    freeze host
    failover host
    failback host
    attempt to create volume

Actual Results: Can modify backend even though it's frozen.
Expected Results: Backend is frozen and must be thawed.

I think it better is change the failover-host function in manager.py from:

3350 if secondary_backend_id == "default":
3351 service.replication_status =fields.ReplicationStatus.ENABLED
3352 service.active_backend_id = ""
3353 service.disabled = False
3354 service.disabled_reason = ""
3355 service.save()

to:

3350 if secondary_backend_id == "default":
3351 service.replication_status = fields.ReplicationStatus.ENABLED
3352 service.active_backend_id = ""
3353 if service.frozen:
3354 service.disabled = True
3355 service.disabled_reason = "frozen"
3356 else:
3357 service.disabled = False
3358 service.disabled_reason = ""
3359 service.save()

if not change the code, there is another problem too, after failback, the 'frozen' filed in the result of "cinder service-list --withreplicaiton True" will be still True and confuse user.

$ cinder service-list --withreplication True
+------------------+-----------------------------------------------------------------------+------+---------+-------+----------------------------+--------------------+-------------------+--------+-----------------+
| Binary | Host | Zone | Status | State | Updated_at | Replication Status | Active Backend ID | Frozen | Disabled Reason |
+------------------+-----------------------------------------------------------------------+------+---------+-------+----------------------------+--------------------+-------------------+--------+-----------------+
| cinder-backup | charmander | nova | enabled | up | 2016-05-02T14:23:28.000000 | | | | - |
| cinder-scheduler | charmander | nova | enabled | up | 2016-05-02T14:23:26.000000 | | | | - |
| cinder-volume | charmander@IBM-DS8K_9.11.242.187_P7_fibre_channel | nova | enabled | up | 2016-05-02T14:23:34.000000 | enabled | | True | |

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/312334

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/312334
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=1f0f01a7c176e095991d95de6ad179b0d45923dd
Submitter: Jenkins
Branch: master

commit 1f0f01a7c176e095991d95de6ad179b0d45923dd
Author: jiamin <email address hidden>
Date: Tue May 3 23:42:16 2016 -0700

    Failback will enable the host that has been frozen

    If user freezes the host before failover, the cinder volume serivce
    will be enabled by failback, and user doesn't need to thaw the host
    before modify the host, if this is expected, the 'Frozen' field in
    cinder service list should be changed to 'False', but it may be better
    that host that is frozen must be thawed.

    Change-Id: Iac03faa9c6cf96531db22e667bd9c194dff48221
    Closes-Bug: #1577417

[Doug Hellmann (doug-hellmann)]

This issue was fixed in the openstack/cinder 9.0.0.0b1 development milestone.