yahoo-eng-team team mailing list archive

[Anthony Lee <anthony.mic.lee@xxxxxx>]

Public bug reported:

There is currently a Cinder spec in-progress that proposes the removal
of file locks that are present during volume attach / detach and a few
other locations. Nova does not appear to be handling exceptions thrown
during the volume detach process in a way that notifies the user why the
detach failed.

Example of what happens when an exception is thrown during a detach by
Cinder's API:

http://paste.openstack.org/show/KBpPWxfVMmQ5GmLeAFpG/

Related Cinder WIP spec giving an overview of why Cinder API might throw
exceptions now:

https://review.openstack.org/#/c/149894/

Related Cinder WIP patch showing potential changes to Cinder API:

https://review.openstack.org/#/c/153748/

When a volume is in an 'ING' state, Cinder API calls that interact with
that volume will return an exception notifying a caller that that volume
is busy. There may be other calls to the Cinder API (that deal with
volumes) that Nova does not handle an exception from. Exceptions from
those calls will need to be handled, too.

In order to reproduce the above exception:

Add 'raise exception.VolumeIsBusy(message="sample")'  to the top of the begin_detaching function in cinder/api.py.
restart c-api.
After attaching a volume in OpenStack, attempt to detach it. The above exception will occur in n-api.

** Affects: nova
     Importance: Undecided
         Status: New

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1458958

Title:
  Exceptions from Cinder detach volume API not handled

Status in OpenStack Compute (Nova):
  New

Bug description:
  There is currently a Cinder spec in-progress that proposes the removal
  of file locks that are present during volume attach / detach and a few
  other locations. Nova does not appear to be handling exceptions thrown
  during the volume detach process in a way that notifies the user why
  the detach failed.

  Example of what happens when an exception is thrown during a detach by
  Cinder's API:

  http://paste.openstack.org/show/KBpPWxfVMmQ5GmLeAFpG/

  Related Cinder WIP spec giving an overview of why Cinder API might
  throw exceptions now:

  https://review.openstack.org/#/c/149894/

  Related Cinder WIP patch showing potential changes to Cinder API:

  https://review.openstack.org/#/c/153748/

  When a volume is in an 'ING' state, Cinder API calls that interact
  with that volume will return an exception notifying a caller that that
  volume is busy. There may be other calls to the Cinder API (that deal
  with volumes) that Nova does not handle an exception from. Exceptions
  from those calls will need to be handled, too.

  In order to reproduce the above exception:

  Add 'raise exception.VolumeIsBusy(message="sample")'  to the top of the begin_detaching function in cinder/api.py.
  restart c-api.
  After attaching a volume in OpenStack, attempt to detach it. The above exception will occur in n-api.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1458958/+subscriptions