openstack
/
cinder
Code Issues Proposed changes

Fix attachment_delete for reserved case 

In attachment-delete we attempt a little short cut
in cases where the volume is reserved but not actually
attached to an instance. In this case we don't have a
connector and we never actually sent anything to the
driver, so we take a short cut in cinder.volume.api
and we just call an object.destroy().

The problem with this is that the object doesn't do
the proper/needed context settings and checks and just
fails due to policy. We should clean this up and add
the extra db update steps and also include the
notification messages.

Change-Id: Ib448f0548739cd42edf03eab07f305b3ac184d8f
Closes-Bug: #1702512
This commit is contained in:
John Griffith 2017-07-05 16:06:55 +00:00
parent 640b870f9c
commit 80b943c7f2
1 changed files with 6 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
cinder/volume/api.py
View File

@ -2003,7 +2003,12 @@ class API(base.Base):
def attachment_delete(self, ctxt, attachment):
volume = objects.Volume.get_by_id(ctxt, attachment.volume_id)
if attachment.attach_status == 'reserved':
attachment.destroy()
self.db.volume_detached(ctxt.elevated(), attachment.volume_id,
attachment.get('id'))
self.db.volume_admin_metadata_delete(ctxt.elevated(),
attachment.volume_id,
'attached_mode')
volume_utils.notify_about_volume_usage(ctxt, volume, "detach.end")
else:
self.volume_rpcapi.attachment_delete(ctxt,
attachment.id,