libvirt 0.7.7 and Lucid

Dustin Kirkland kirkland at canonical.com
Fri Apr 2 03:29:11 UTC 2010


On Thu, Apr 1, 2010 at 12:01 PM, Scott Moser <smoser at ubuntu.com> wrote:
> On Wed, 31 Mar 2010, Dustin Kirkland wrote:
>> On Thu, Mar 25, 2010 at 1:09 PM, Jamie Strandboge <jamie at canonical.com> wrote:
>> > 3. libvirt seems to try to guard against hypervisor issues regarding
>> > detach-device and detach-disk. Specifically, these can be hot-plugged,
>> > but not hot-unplugged (libvirt claims the qemu-kvm hypervisor doesn't
>> > support detach of these):
>> >    <disk type='block' device='disk'>
>> >      <driver name='phy'/>
>> >      <source dev='...path.../device_disk.img'/>
>> >      <target dev='sdb' bus='scsi'/>
>> >      <alias name='scsi0-0-1'/>
>> >      <address type='drive' controller='0' bus='0' unit='1'/>
>> >    </disk>
>
> hotplug attach and detach of scsi driver disks is an absolute requirement
> for UEC with Eucalyptus 1.6.2.  It is the basis of the "EBS" (Elastic
> Block Store) functionality (euca-attach-volume, euca-detach-volume).

Jamie and I spent a good bit of today testing libvirt 0.7.7, in
particular against Lucid UEC.

The basic tests from Mathias' lp:~mathiaz/%2Bjunk/uec-testing-scripts
all ran just fine, as well as my basic tests of virsh, and
virt-manager.

The package build also runs the upstream test suite, all test of which pass:
 * http://launchpadlibrarian.net/42047543/buildlog_ubuntu-lucid-amd64.libvirt_0.7.7-4ubuntu1~jdstrand2_FULLYBUILT.txt.gz

However, hot add of scsi disks (as used by Eucalyptus's EBS
implementation) did not work, unfortunately.  Note that we used these
instructions from Scott to test:
 * http://paste.ubuntu.com/407717/

After a conversation with upstream in #virt in IRC, Jamie filed the
upstream bug:
 * https://bugzilla.redhat.com/show_bug.cgi?id=578975

We think that we've identified the problematic commit (but this
doesn't unapply cleanly).  See:
264e98d6a85bb467f364006c1dac3993688691ce [PATCH] Make hotplug use new
device_add where possible

Libvirt needs to go back to using pci_add for scsi hot attach of disks.

Jamie and I are in agreement that this regression is a blocker for
uploading 0.7.7.

We also defined the functional acceptance criteria for 0.7.7 upload to
be (assuming we can get this fixed):
 * the upstream QA tests all pass
 * the security regression tests all pass
 * Mathias' UEC tests all pass
 * and Dan's S3/EBS tests all pass

:-Dustin




More information about the ubuntu-server mailing list