yahoo-eng-team team mailing list archive

[OpenStack Infra <1200113@xxxxxxxxxxxxxxxxxx>]

Reviewed:  https://review.openstack.org/50932
Committed: http://github.com/openstack/nova/commit/19e2b7ddfc7035565e75444e5813a9d195c85bfb
Submitter: Jenkins
Branch:    milestone-proposed

commit 19e2b7ddfc7035565e75444e5813a9d195c85bfb
Author: Jiajun Liu <jiajun@xxxxxxxxxxxxxxx>
Date:   Thu Aug 22 11:45:18 2013 +0800

    fallocate image only when user has write access
    
    Libvirt will change instance's ownership when the instance became power
    off which will make nova-compute service unable to hard reboot instance
    bacause the nova-compute service has not write access. This commit will
    allocate space for image only the first time we create image.
    
    fixes bug 1200113
    
    Change-Id: Id81b09ade5f870665f5ffabab3cde144c6dec210
    (cherry picked from commit c246fb37e20197be27264a36515725233085df1d)


** 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/1200113

Title:
  hard reboot fails with preallocate_images=performance

Status in OpenStack Compute (Nova):
  Fix Released

Bug description:
  I have in /etc/nova/nova.conf

  preallocate_images=performance

  for obvious speed reasons.

  I had the need to hard reboot some instances (somehow soft reboot does
  not seem to work often with me) and I got this error in the logs for
  every instance:

  013-07-10 18:53:40.638 DEBUG nova.utils [req-911ae4d0-6a75-416f-b3d7-b281793cbfdc 93178802ad6f48e8aea05cc7ba195859 876e74dbf00f44418c6a61d83c5b7ce8] Running cmd (subprocess): fallocate -n -l 42949672960 /var/lib/nova/instances/abbb258d-f851-4f4e-81ee-0b8334d331d2/disk execute /usr/lib/python2.7/dist-packages/nova/utils.py:208
  2013-07-10 18:53:40.652 DEBUG nova.utils [req-911ae4d0-6a75-416f-b3d7-b281793cbfdc 93178802ad6f48e8aea05cc7ba195859 876e74dbf00f44418c6a61d83c5b7ce8] Result was 1 execute /usr/lib/python2.7/dist-packages/nova/utils.py:232
  2013-07-10 18:53:40.653 ERROR nova.compute.manager [req-911ae4d0-6a75-416f-b3d7-b281793cbfdc 93178802ad6f48e8aea05cc7ba195859 876e74dbf00f44418c6a61d83c5b7ce8] [instance: abbb258d-f851-4f4e-81ee-0b8334d331d2] Cannot reboot instance: Unexpected error while running command.
  Command: fallocate -n -l 42949672960 /var/lib/nova/instances/abbb258d-f851-4f4e-81ee-0b8334d331d2/disk
  Exit code: 1
  Stdout: ''
  Stderr: 'fallocate: /var/lib/nova/instances/abbb258d-f851-4f4e-81ee-0b8334d331d2/disk: open failed: Permission denied\n'

  Basically the hard reboot tries to re-allocate the disk space (which -
  in this case? - is already allocated) and fails because the disk file
  is owned by root (I think) rather than nova. Doing a chown nova:nova
  on the disk files resolves the issue for me.

  I see three (combined) options:

  1) do not try to reallocate in case it is already the right size
  2) execute the fallocate part under root
  3) make sure the disk is owned/writeable by nova

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1200113/+subscriptions