[Bug 2165345] Re: [SRU] Allow 4k aligned IO for windows guests

[Zachary Raines]

I've also created ppas with test packages:

 https://launchpad.net/~raineszm/+archive/ubuntu/lp2165345-testing-ae
 https://launchpad.net/~raineszm/+archive/ubuntu/lp2165345-testing-bf
 https://launchpad.net/~raineszm/+archive/ubuntu/lp2165345-testing-nj
 https://launchpad.net/~raineszm/+archive/ubuntu/lp2165345-testing-yc
 https://launchpad.net/~raineszm/+archive/ubuntu/lp2165345-testing-zd

-- 
You received this bug notification because you are a member of Ubuntu
OpenStack, which is subscribed to cinder in Ubuntu.
https://bugs.launchpad.net/bugs/2165345

Title:
  [SRU] Allow 4k aligned IO for windows guests

Status in Ubuntu Cloud Archive:
  In Progress
Status in Ubuntu Cloud Archive antelope series:
  In Progress
Status in Ubuntu Cloud Archive bobcat series:
  In Progress
Status in Ubuntu Cloud Archive caracal series:
  In Progress
Status in Ubuntu Cloud Archive dalmatian series:
  In Progress
Status in Ubuntu Cloud Archive epoxy series:
  In Progress
Status in Ubuntu Cloud Archive flamingo series:
  In Progress
Status in Ubuntu Cloud Archive yoga series:
  In Progress
Status in Ubuntu Cloud Archive zed series:
  In Progress
Status in cinder package in Ubuntu:
  In Progress
Status in cinder source package in Jammy:
  In Progress
Status in cinder source package in Noble:
  In Progress

Bug description:
  [ Impact ]

  Ceph performs much better when I/O is 4k-aligned otherwise it has to
  read the 4k from disk, modify and write it again. Linux guests
  generally submit 4k-aligned I/O, however Windows guests generally
  submit 512b-aligned I/O. This can lead to significantly degraded
  performance for Openstack Windows guests when using the Ceph RBD
  backend.

  This SRU backports upstream patches[0-2] which enable hinting the
  logical block size of cinder devices to Nova and which in turn allows
  the Windows guest to perform 4k-aligned I/O.

  [ Test Plan ]

  1. Deploy openstack with cinder enabled and the ceph backend

  2. Create a new volume and ensure that the blocksize is not passed to
  libvirt

  3. Create a new volume with the disk_geometry option and ensure that
  the block size is properly passed to libvirt

  [ Where problems could occur ]

  * Virtual disks may not support setting block size [0]. As a result, setting the disk_geometry on LUN drives may cause errors in Nova. This is mitigated by leaving the disk_geometry option unset by default and allowing users to opt in on a driver by driver basis.
      * This has already been surfaced and addressed by the second patch of the series [2].

  [ Other Info ]

  * As noted in [0], for existing volumes, the libvirt xml will need to be regenerated for the effects to be seen e.g., by reattaching the volume. See the release-notes for the patch [1] for more details.
      * A corollary of the above, after some discussion with @mruffell, its probably best to backport this to interim releases as well as LTS, to minimize the number of times this detach/re-attach process needs to be carried out.

  [0]: https://bugs.launchpad.net/bugs/2127196
  [1]: https://review.opendev.org/c/openstack/cinder/+/658283
  [2]: https://review.opendev.org/c/openstack/cinder/+/975540

To manage notifications about this bug go to:
https://bugs.launchpad.net/cloud-archive/+bug/2165345/+subscriptions