[libvirt] Issues with qemu-nbd over AF_UNIX and virDomainAttachDevice

[Deepak C Shetty]

On 05/27/2013 04:41 PM, Paolo Bonzini wrote:
> Il 27/05/2013 12:46, Deepak C Shetty ha scritto:
>> On 05/27/2013 02:00 PM, Paolo Bonzini wrote:
>>> Il 26/05/2013 15:54, Deepak C Shetty ha scritto:
>>>> 3) In nbdxml, i tried removing port="1111".. it didn't give any error
>>>> with regards to port beign mandatory.. which si good.. but the attach
>>>> device still gives the same error as above
>>>>
>>>> Is the `nbdxml` I am using for attachign a qemu-nbd exported drive,
>>>> correct ?
>>> What version of QEMU is this?
>>>
>>> Can you search the logs for the QMP commands related to the hotplug?
>> Do you mean starting my domain using -d option for qemu, which dumps the
>> log in /tmp/qemu.log ?
> I mean libvirt.log.  You can start libvirtd with these environment
> variables:
>
> LIBVIRT_DEBUG=debug LIBVIRT_LOG_OUTPUTS=1:file:/tmp/libvirt.log

Thanks, this helped. Here is the summary

With latest libvirtd (1.0.5 built from git)...

<disk type='network'>
   <driver name="qemu" type="qcow2"/>
   <source protocol="nbd">
     <host name="deepakcs-lx" transport="unix" socket="/tmp/mysock2" />
   </source>
   <target dev="vdc" bus="virtio" />
</disk>

 >>> dom.attachDevice(nbdxml)
libvirt: QEMU Driver error : operation failed: open disk image file failed
Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
   File "/home/dpkshetty/work/newlibvirt2/libvirt/python/libvirt.py", 
line 419, in attachDevice
     if ret == -1: raise libvirtError ('virDomainAttachDevice() failed', 
dom=self)
libvirt.libvirtError: operation failed: open disk image file failed

* from /tmp/libvirt.log
QEMU_MONITOR_IO_WRITE: mon=0x7f8308003d40 
buf={"execute":"human-monitor-command","arguments":{"command-line":"drive_add 
dummy 
file=nbd:unix:/tmp/mysock2,if=none,id=drive-virtio-disk2,format=qcow2"},"id":"libvirt-390"}

QEMU_MONITOR_RECV_REPLY: mon=0x7f8308003d40 reply={"return": "could not 
open disk image nbd:unix:/tmp/mysock2: Invalid argument\r\n", "id": 
"libvirt-390"}

On command line...
qemu-kvm -drive file=nbd:unix:/tmp/mysock2,format=qcow2
qemu-kvm: -drive file=nbd:unix:/tmp/mysock2,format=qcow2: could not open 
disk image nbd:unix:/tmp/mysock2: Invalid argument

but

qemu-kvm -drive file=nbd:unix:/tmp/mysock2 - works!

So looks like format=qcow2 is causing issues!!!

Then tried...
(removing the <driver... altoghether)

<disk type='network'>
   <source protocol="nbd">
     <host name="deepakcs-lx" transport="unix" socket="/tmp/mysock2" />
   </source>
   <target dev="vdc" bus="virtio" />
</disk>

 >>> dom.attachDevice(nbdxml)
0

 >>> dom.detachDevice(nbdxml)
0
 >>>


* from /tmp/libvirt.log
QEMU_MONITOR_IO_WRITE: mon=0x7f8308003d40 
buf={"execute":"human-monitor-command","arguments":{"command-line":"drive_add 
dummy 
file=nbd:unix:/tmp/mysock2,if=none,id=drive-virtio-disk2"},"id":"libvirt-867"}

This works and I was able to successfully add/remove a disk exported via 
qemu-nbd to a running VM !


>
>
>> I am using a VM started from virt-manager.... i don't see a way to pass
>> -d to it from virt-manager... I can try using a hand-coded qemu cmdline
>> tho'
>>
>> I assume when i am using python (import libvirt) and/or virsh.. it uses
>> qemu-kvm
>> qemu-kvm version is 1.0.1
> Ok.
>
>> Do i need to try with latest qemu git ?
> No, I don't think so, but you can try (1.4.0 or 1.5.0 will do).

I was using qemu-kvm version 1.0.1... for all of the ^^ above

>
>> On irc.. you asked me to check if cold-plug works.. but attach-device is
>> for active domain only... am i missing something here ?
> You can try putting the <disk> item and start the VM.  In this case the
> logs in /var/log/libvirt/qemu will be helpful, because they contain the
> command-line that is used to start QEMU.

Tried putting the above nbdxml usign virsh edit <domname> as an addnl 
disk and domain booted fine
It throws the same error if u add format=qcow2 under <driver>...

So looks like the right way to use NBD is *not* to specify format and 
let QEMU autosense it ?

thanx,
deepak


>
> Paolo
>
>