yahoo-eng-team team mailing list archive

[Rongze Zhu <zrzhit@xxxxxxxxx>]

Public bug reported:

The Reproduce steps:

I create a volume from image(the cinder backend is HDS FC Storage).

Then I boot a instance from the volume, the call path is "nova-api => nova-compute => cinder-api => cinder-volume".
Because the HDS Driver in cinder-volume is slow in my env, so nova-compute will timeout and set the status of instance as "error".

Then I try to delete the instance, because the value of 'os-extended-
volumes:volumes_attached' in the instance is the volume id, so nova-
compute will call detach-volume API in cinder-api.  Because the status
of volume in cinder database is 'available', so the cinder-api will
raise a 'VolumeAttachmentNotFound' exception. Finally, I failed to
delete the error instance.

Solution：

nova-compute need to check the status of volume before call detach-
volume api to cinder-api.

** Affects: nova
     Importance: Undecided
         Status: New

** Description changed:

  The Reproduce steps:
  
  I create a volume from image(the cinder backend is HDS FC Storage).
  
  Then I boot a instance from the volume, the call path is "nova-api => nova-compute => cinder-api => cinder-volume".
- Because the HDS Driver in cinder-volume is slow in my env, so nova-compute will timeout and set the stat of instance as "error".
+ Because the HDS Driver in cinder-volume is slow in my env, so nova-compute will timeout and set the status of instance as "error".
  
  Then I try to delete the instance, because the value of 'os-extended-
  volumes:volumes_attached' in the instance is the volume id, so nova-
- compute will call detach-volume API in cinder-api.  Because the state of
- volume in cinder database is 'available', so the cinder-api will raise a
- 'VolumeAttachmentNotFound' exception. Finally, I failed to delete the
- error instance.
- 
+ compute will call detach-volume API in cinder-api.  Because the status
+ of volume in cinder database is 'available', so the cinder-api will
+ raise a 'VolumeAttachmentNotFound' exception. Finally, I failed to
+ delete the error instance.
  
  Solution：
  
- nova-compute need to check the state of volume before call detach-volume
- api to cinder-api.
+ nova-compute need to check the status of volume before call detach-
+ volume api to cinder-api.

-- 
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/1552265

Title:
  can't delete a error instance(boot from volume )

Status in OpenStack Compute (nova):
  New

Bug description:
  The Reproduce steps:

  I create a volume from image(the cinder backend is HDS FC Storage).

  Then I boot a instance from the volume, the call path is "nova-api => nova-compute => cinder-api => cinder-volume".
  Because the HDS Driver in cinder-volume is slow in my env, so nova-compute will timeout and set the status of instance as "error".

  Then I try to delete the instance, because the value of 'os-extended-
  volumes:volumes_attached' in the instance is the volume id, so nova-
  compute will call detach-volume API in cinder-api.  Because the status
  of volume in cinder database is 'available', so the cinder-api will
  raise a 'VolumeAttachmentNotFound' exception. Finally, I failed to
  delete the error instance.

  Solution：

  nova-compute need to check the status of volume before call detach-
  volume api to cinder-api.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1552265/+subscriptions