Guo Lian Yun wrote:
libvirt-cim-bounces@redhat.com wrote on 2009-03-14 03:01:14:
+ drasd = get_typed_class(options.virt, 'DiskResourceAllocationSettingData') + drasd_id = '%s/%s' % (default_dom, nddev) + dkeys = {'InstanceID' : drasd_id} + + nrasd = get_typed_class(options.virt, 'NetResourceAllocationSettingData') + nrasd_id = '%s/%s' % (default_dom, nmac) + nkeys = {'InstanceID' : nrasd_id}
Instead of hand building the instances for the resources, use the proper association calls to get the RASD instances. You know the guest, so you can use SystemDevice to get the resource instance, and then SettingsDefineState to get the RASD instances.
I'm working on a new patch to get the RASD instances by association. But I can't remove "KVM_InputResourceAllocationSettingData" which output the device "mouse:ps2" can not be found. Actually, I get it by SettingsDefineState. If I remove this device from the ResourceSettings, this method works for me.
Sorry Daisy.. I'm not sure I understand the problem you're hitting.
Are you seeing an error trying to remove an input device from a guest? What error message do you see?
Sorry for my confusion. I send out #2 patch. The problem is that I can not remove InputRASD. Below is the report error:
VirtualSystemManagementService - 16_removeresource.py: FAIL ERROR - 'CIMInstanceName' object has no attribute 'DeviceID' No such instance (no device domain/mouse:ps2)
Which virtualization type are you seeing this issue on? I'm not seeing an issue with KVM. I tried running on a Xen system, and I get the following error: VirtualSystemManagementService - 16_removeresource.py: FAIL ERROR - Got CIM error CIM_ERR_FAILED: Failed to define domain: invalid argument in virDomainDefineXML with return code 1 ERROR - Failed to define the dom: domain InvokeMethod(DefineSystem): CIM_ERR_FAILED: Failed to define domain: invalid argument in virDomainDefineXML Can you paste the provider debug you're seeing?
So I Skip InputRASD together with MemRASD and ProcRASD in #2 patch.
The providers should be able to remove input devices, so you'll want to include that in the test. If you're seeing an error, it's either a test case issue or a provider issue. It's better to resolve these issues before resending an updated patch (it'll reduce the amount of churn in a given patch). Thanks! -- Kaitlin Rupert IBM Linux Technology Center kaitlin@linux.vnet.ibm.com