[Libvirt-cim] [PATCH] [TEST] Fixing vsms/08_modifyresource.py

[Deepti B Kalakeri]

Kaitlin Rupert wrote:
> Deepti B Kalakeri wrote:
>>
>>
>> Kaitlin Rupert wrote:
>>> Deepti B. Kalakeri wrote:
>>>> # HG changeset patch
>>>> # User Deepti B. Kalakeri<deeptik at linux.vnet.ibm.com
>>>> # Date 1260446374 28800
>>>> # Node ID e95a3d156fc2284484f98f07c39fbfd65929a395
>>>> # Parent 683a8a57d54506032a659bdb9178b7558827ffde
>>>> [TEST] Fixing vsms/08_modifyresource.py
>>>>
>>>> Tested with KVM and current sources on Fedora
>>>> Signed-off-by: Deepti B. Kalakeri <deeptik at linux.vnet.ibm.com>
>>>>
>>>> diff -r 683a8a57d545 -r e95a3d156fc2 
>>>> suites/libvirt-cim/lib/XenKvmLib/vsms_util.py
>>>> --- a/suites/libvirt-cim/lib/XenKvmLib/vsms_util.py Sat Dec 05 
>>>> 02:22:09 2009 -0200
>>>> +++ b/suites/libvirt-cim/lib/XenKvmLib/vsms_util.py Thu Dec 10 
>>>> 03:59:34 2009 -0800
>>>> @@ -120,9 +120,10 @@
>>>> cxml.dumpxml(server)
>>>> dom = cxml.xml_get_dom_name()
>>>> cpu = virsh_vcpuinfo(server, dom, virt)
>>>> - if cpu is None:
>>>> + if int(cpu) < int(ncpu) or cpu is None:
>>>
>>> This works, but it's more of a work around. In this case, vcpuinfo 
>>> returns an error. When we check the output for the VCPU string, we 
>>> get zero occurrences, which is what virsh_vcpuinfo() returns.
>>>
>>> A better check would be to see if the virsh call returns an error. 
>>> If it does, virsh_vcpuinfo() should return None. Otherwise, it 
>>> should return an integer.
>> virsh_vcpuinfo() uses the following command cmd = "virsh -c %s 
>> vcpuinfo %s 2>/dev/null | grep VCPU | wc -l" % (virt2uri(virt), dom)
>> which would return success even when no records are found. So I think a 
>
> Agreed. I think that's a flaw of the virsh_vcpuinfo() call though. 
> Because you never know whether "virsh -c %s vcpuinfo %s" fails. I 
> would remove the "| grep VCPU | wc -l" portion of the command. Check 
> to see if the call fails - if it does, return None from 
> virsh_vcpuinfo(). If the call doesn't fall, parse the output to get 
> the VCPU count.
>
Yes! This is what I was thinking of, but was as you said this will take 
little more code than the one line which did the work for us.
> Thoughts? This is a lot more work, but then you would know when the 
> virsh call is failing (as opposed to just returning 0 VCPus).
I will work on this and send the patch.
>
>> better way would be to verify the value returned and return the out 
>> when command returns something >0,
>> otherwise None.
>>
>> Any suggestions ?
>
>
>>
>
>

-- 
Thanks and Regards,
Deepti B. Kalakeri
IBM Linux Technology Center
deeptik at linux.vnet.ibm.com