xenapi: issues with destroy_vdi masking real errors

[John Garbutt]

In cleaning up VDIs we call destroy_vdi.

However, should the destroy fail, it masks the real error, for example:

Unable to destroy VDI OpaqueRef:f53f35a9-31e1-af85-87a8-aa0dc7eecd43
  nova/compute/manager.py", _build_instance
...
  nova/virt/xenapi/vmops.py", line, in _attach_disks
    DEVICE_SWAP, name_label, swap_mb)
  nova/virt/xenapi/vm_utils.py", line, in generate_swap
    'swap', swap_mb, fs_type)
 nova/virt/xenapi/vm_utils.py", line, in _generate_disk
    destroy_vdi(session, vdi_ref)
  nova/virt/xenapi/vm_utils.py", line, in destroy_vdi
    _('Unable to destroy VDI %s') % vdi_ref)

We should instead use the safe_destroy_vdi call.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/54526

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/54526
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=04e53669dd3e6acd8086fd06ea073a02207ffb24
Submitter: Jenkins
Branch: master

commit 04e53669dd3e6acd8086fd06ea073a02207ffb24
Author: John Garbutt <email address hidden>
Date: Wed Oct 30 11:38:24 2013 +0000

    xenapi: stop destroy_vdi errors masking real error

    We frequently call destroy_vdi in an exception block where we plan to
    re-raise the exception.

    Should an error occur while destroying the vdi this hides the real error
    that caused the problem. To stop this, I have moved those calls to
    safe_destroy_vdis and added unit tests to ensure the error from
    destroy_vdi does not mask the real issue.

    Closes-Bug: #1246276
    Change-Id: Ib02f93ab1086f077d3d00e5ca915a26f90552c85