Re: [Libvirt-cim] KVM on Pegasus Test Run Summary for Dec 15 2008
> FAIL Test Summary:
> VirtualSystemManagementService - 14_define_sys_disk.py: FAIL
The above tc is failing because of the the missing VirtualDevice field
in the DiskRASD passed to DefineSystem().
The net, disk, proc, mem RASD properties passed to the DefineSystem() in
this tc are obtained by the association of AC with
SettingsDefineCapabilities.
With the F9 rpm the VirtualDevice field in the DiskRASD is not set and
hence the DefineSystem() throws an error.
The tc passes when we include rasd['VirtualDevice'] field to contain
appropriate VirtualDevice value 'hda' in case [KVM] of F9 rpm.
I have a doubt here, the changes to throw an error on Missing
VirtualDevice field was included in the modifications made to the patch
with revision 681
"Remove the default implied VirtualDevice on DiskRASD", but the changes
included in the F9 rpm are upto 613.
Can you let me know upto which changeset does the F9 rpm cover?
This is because the F9 rpm has additional patches applied to it. The
last rpm update we did, we added the DiskRASD patch (along with a few
others).
You can tell patches have been applied, because the revision number is
613+ instead of 613.
This makes branching the test difficult because you cannot rely on the
revision number alone.
What you can do is check the changeset value. In const.py, you can
designate the change set as the F9 changeset. This will allow you to
branch tests appropriately.
> VirtualSystemManagementService - 15_mod_system_settings.py: FAIL
The tc fails only with the KVM F9 rpm.
The value of AutomaticRecoveryAction is set to 2 only for KVM with F9
rpm, otherwise the value is 3 for Xen and KVM with current sources.
I am not able to track if this was expected behavior of setting
AutomaticRecoveryAction to 2 when the F9 rpm was built and was changed
later to be set to 3 for the modifications beyond revision 613.
Can you help ?
Looks like this is being caused by a bug in the F9 libvirt rpm. Do the
following to reproduce this issue just using virsh:
1) Run the test, dump the guest xml.
2) Modify the guest xml so that <on_crash> is set to restart.
3) Define the guest using the xml.
4) Dump the xml of the newly defined guest. The value for <on_crash>
will be destroy, not restart.
I'd add a bug on the wiki page and have this test XFAIL on F9.
--
Kaitlin Rupert
IBM Linux Technology Center
kaitlin(a)linux.vnet.ibm.com