yahoo-eng-team team mailing list archive

[Matt Rabe <mdrabe@xxxxxxxxxx>]

Public bug reported:

Description
===========
If a recreate fails after the instance's volumes have their connections initialized on the destination, the connections are not terminated.

Steps to reproduce
==================
During the recreate flow, if after the volumes are attached to the destination something fails, then the volumes are left attached.

To reproduce one could patch manager.py on the destination host and if
the compute driver uses the default manager rebuild implementation then
force an exception to be raised after the attach_block_devices call in
_rebuild_default_impl. Then issue an evacuation of an instance to this
host after picking up the patch.

Expected result
===============
Storage cleanup needs to occur, e.g. host mappings on the storage provider for the destination host need to be removed.

Actual result
=============
Host mappings remain on the storage provider.

Environment
===========
1. Exact version of OpenStack you are running. See the following
   Newton

2. Which hypervisor did you use?
   PowerVM (nova-powervm)

2. Which storage type did you use?
   Cinder (storwize)

3. Which networking type did you use?
   Neutron with SEA

** 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/1668432

Title:
  Failed evacuations leave cinder volumes mapped to destination host

Status in OpenStack Compute (nova):
  New

Bug description:
  Description
  ===========
  If a recreate fails after the instance's volumes have their connections initialized on the destination, the connections are not terminated.

  Steps to reproduce
  ==================
  During the recreate flow, if after the volumes are attached to the destination something fails, then the volumes are left attached.

  To reproduce one could patch manager.py on the destination host and if
  the compute driver uses the default manager rebuild implementation
  then force an exception to be raised after the attach_block_devices
  call in _rebuild_default_impl. Then issue an evacuation of an instance
  to this host after picking up the patch.

  Expected result
  ===============
  Storage cleanup needs to occur, e.g. host mappings on the storage provider for the destination host need to be removed.

  Actual result
  =============
  Host mappings remain on the storage provider.

  Environment
  ===========
  1. Exact version of OpenStack you are running. See the following
     Newton

  2. Which hypervisor did you use?
     PowerVM (nova-powervm)

  2. Which storage type did you use?
     Cinder (storwize)

  3. Which networking type did you use?
     Neutron with SEA

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1668432/+subscriptions