yahoo-eng-team team mailing list archive

[Thierry Carrez <thierry.carrez+lp@xxxxxxxxx>]

** Changed in: nova
       Status: Fix Committed => Fix Released

-- 
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/1224918

Title:
  Don't generate notifications when reaping running_deleted instances

Status in OpenStack Compute (Nova):
  Fix Released

Bug description:
  The periodic task which looks for instances which have been deleted in
  the DB but are still running at the hypervisor supports a "reap"
  action to delete them.  It does this by calling the
  _shutdown_instance() method, which as part of its processing will send
  notification messages.

  However, the notification code depends on the instance object
  including system metadata information, and the db queries that build
  instance objects do not extend the "read deleted" filter down to these
  queries.   This results in an exception in the shutdown code, which in
  turn prevents the  VM from being deleted.

  
  Feels like there are three issues here:

  i) The reap code should not send notification messages.  The instance
  was deleted some time ago, and so messages will have been generated
  then.  Also there will have been no "exists"  messages for this
  instance.  Sending new messages for the instance at this point will
  confuse any billing systems looking for such messages.

  ii) The notification code should be more robust and use get() methods
  when reading data from the instance object

  iii) The DB queries for building an instance should carry the "read
  deleted" filter into all sub queries for consistency.

  
  This bug relates to just the first of these - other bugs will be filed for the other two points.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1224918/+subscriptions