[virt-tools-list] Virt-manager doesn't attach Sheepdog storage VDI disk correctly

Cole Robinson crobinso at redhat.com
Tue Aug 11 14:02:55 UTC 2015

On 08/11/2015 12:13 AM, Dmitriy Slachshyov wrote:
> So, I represented everything that I did before. I've got unsuccessful
> results. The xml still almost the same as early.
> https://github.com/sheepdog/sheepdog/wiki/Libvirt
> This is a sheepdog maintainer's web site. There is a detailed
> description of format for libvirtd. I have tested it before and it
> works right.
> Also look at output of commands, I 've got after the test was finished:
> # virsh pool-list
>  Name                 State      Autostart 
> -------------------------------------------
>  default              active     yes       
>  LVM                  active     yes       
>  sheepdog             active     yes       
>  virt-manager         active     yes       
> # virsh vol-list sheepdog
>  Name                 Path                                    
> -------------------------------------------
>  dogtest2             dogtest2                                
> # virsh vol-dumpxml dogtest2 sheepdog
> <volume type='network'>
>   <name>dogtest2</name>
>   <key>sheepdog/dogtest2</key>
>   <source>
>   </source>
>   <capacity unit='bytes'>8589934592</capacity>
>   <allocation unit='bytes'>0</allocation>
>   <target>
>     <path>dogtest2</path>                 <--- the mistake is here
>     <format type='unknown'/>
>   </target>
> </volume>
> #

Can you provide XML for a _working_ volume after it's been defined?

The example XML from the link you provide, which is what we have in our
testsuite, has some confusing bits:

- If specifies <key> but that's a readonly field. So I'm interested to see
what libvirt outputs for a working volume
- The volume path is 'sheepdog:myvol'. Is 'sheepdog:' just a protocol
specified, like http://, or does it reference the pool name?
- Also, from looking at the libvirt code, it seems <target><path> is readonly
as well, so whatever virt-manager specifies is just thrown out. So I'm curious
to see what libvirt spits out for a working volume

Also please put all this in a bug report. Starting tomorrow I'm offline for
two weeks, I want to make sure this info doesn't get lost in the mail shuffle
when I come back


> On Mon, 10/08/2015 at 17:06 -0400, Cole Robinson wrote:
>> On 08/10/2015 06:08 AM, Dmitriy Slachshyov wrote:
>>> Cole,
>>> I've repeated the adding process, running --debug flag.
>>> Look at the attached file with debug messages.
>> Thanks for the info. Unfortunately I don't have a sheepdog setup so I 
>> can't
>> debug this easily at the moment. However I pushed a virt-manager 
>> commit that
>> will give more debugging info. Can you do the following:
>> git clone git://fedorahosted.org/virt-manager.git
>> cd virt-manager
>> ./virt-manager --debug
>> Reproduce the issue again: creating a volume in virt-manager, then 
>> selecting
>> it, and trying to continue the install.
>> Take all that info and the summary of your original mail and put it 
>> in a bug
>> report. https://virt-manager.org/bugs/
>> Also please provide 'virsh vol-dumpxml --pool $POOLNAME $VOLNAME' for 
>> a
>> correctly created sheepdog volume
>> Thanks,
>> Cole

More information about the virt-tools-list mailing list