Attaching the same multiattach volume to the same instance twice leaves the volume in 'Attaching' state

[Shyama]

Calling reserve_volume and the attach api in sequence two times for the same instance and same multiattach volume leaves the volume in attaching state

First time
- reserve_volume sets the volume status to 'Attaching'
- attach api updates db with attachment information and marks the volume as 'In-Use'

Second time
- reserve_volume sets the volume to 'Attaching'
- cinder/volume/manager.py checks if the volume is already attached to the same instance and returns, but the volume is left in 'attaching' state

            if instance_uuid:
                attachment = \
                    self.db.volume_attachment_get_by_instance_uuid(
                        context, volume_id, instance_uuid)
            else:
                attachment = \
                    self.db.volume_attachment_get_by_host(context, volume_id,
                                                          host_name_sanitized)
            if attachment is not None:
                return ==> this return leaves the volume in attaching state

The fix would be to revert the state back to in-use before return
@@ -785,6 +785,11 @@
                     self.db.volume_attachment_get_by_host(context, volume_id,
                                                           host_name_sanitized)
             if attachment is not None:
+ # In the case where volume is already attached to this instance
+ # we need to update the volume status from 'attaching' to
+ # 'in-use'
+ self.db.volume_update(context, volume_id,
+ {'status': 'in-use'})
                 return

             self._notify_about_volume_usage(context, volume,

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/179019

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/179019
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=709949d904c7f1d0e13b41bf6723d71bcf6b5652
Submitter: Jenkins
Branch: master

commit 709949d904c7f1d0e13b41bf6723d71bcf6b5652
Author: Shyama Venugopal <email address hidden>
Date: Thu Apr 30 12:22:27 2015 +0200

    Revert state if attachment already exists

    When attach is called twice for the same volume and instance the
    attach_volume checks if the attachment already exists and returns.
    This leaves the volume in 'Attaching' state. The volume status should
    be reset to 'in-use' if we see the attachment already exists.

    Change-Id: I2a2d28de59af234300082ab4655b0bc10699d54c
    Closes-Bug: #1449980

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/kilo
Review: https://review.openstack.org/179207

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/179207
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=a5908fdbcb2b80fdd7cecc04c82abf7c88d1a75e
Submitter: Jenkins
Branch: stable/kilo

commit a5908fdbcb2b80fdd7cecc04c82abf7c88d1a75e
Author: Shyama Venugopal <email address hidden>
Date: Thu Apr 30 12:22:27 2015 +0200

    Revert state if attachment already exists

    When attach is called twice for the same volume and instance the
    attach_volume checks if the attachment already exists and returns.
    This leaves the volume in 'Attaching' state. The volume status should
    be reset to 'in-use' if we see the attachment already exists.

    Change-Id: I2a2d28de59af234300082ab4655b0bc10699d54c
    Closes-Bug: #1449980
    (cherry picked from commit 709949d904c7f1d0e13b41bf6723d71bcf6b5652)