[ovirt-users] can't import vm from KVM host

Tue Nov 8 11:16:55 UTC 2016

On 08.11.16 10:58, Nelson Lameiras wrote:
> Hi Shahar,
> 
> We try to prioritise vm behaviour predictability over ressources consumption, so "thin provisionning" is not a option for us. "Preallocated" is always selected over default behaviour.
> 
> Nevertheless, while trying to import a KVM vm (from another host), I get a 0 disk count on the VM, which means I do not event get to the point where I can chose allocation policy (usually next screen).
> This is true with or without patch proposed below (assuming it has not changed since sept 29).

I assume that this line is failing:
vol = conn.storageVolLookupByPath(disk['alias'])
it may be the reason that the storage is not part of a pool.
(look at pool-xxx commands via virsh)

If it doesn't work it may be related to a different API that we will need to
implement as Tomas suggested in the gerrit patch.

> 
> Can I give you any more information?
> 
> cordialement, regards, 
> Nelson LAMEIRAS 
> 
> Lyra Network 
> Service Projets et Processus 
> Tel : +33 (0) 5 32 09 09 70 
> 109 rue de l’innovation 
> 31670 Labège - France 
> www.lyra-network.com
> 
> ----- Original Message -----
> From: "Shahar Havivi" <shavivi at redhat.com>
> To: "Nelson Lameiras" <nelson.lameiras at lyra-network.com>, "Tomas Golembiovsky" <tgolembi at redhat.com>
> Cc: "Michal Skrivanek" <michal.skrivanek at redhat.com>, users at ovirt.org
> Sent: Monday, November 7, 2016 1:58:34 PM
> Subject: Re: [ovirt-users] can't import vm from KVM host
> 
> Hi Nelson,
> 
> Sorry for the long wait (I was on PTO).
> I just test the patch with the following disk xml:
> 
> <disk type='block' device='disk'>
>     <driver name='qemu' type='raw' cache='none' io='native'/>
>     <source dev='/dev/disk/by-path/...lun0'/>
>     <target dev='vda' bus='virtio'/>
>     <address type='pci' domain='0x0000' bus='0x00'
>     slot='0x07' function='0x0'/>
> </disk>
> 
> I did manage to see the disk with its actual size,
> I was able to import the VM as well when I choose:
> "Allocation Policy": "Preallocated"
> (not the "Thin Provision" as is the default.
> Currently we need to pre allocated kvm block disk and cannot use the thin
> provision (we may need to disable it in the UI).
> 
> Did you try that?
> It may related to the reason that the dev path you have is a /dev/mapper/ and
> this is the reason that libvirt not returning the disk size? (Tomas?)
> 
>  Shahar.
> 
> 
> On 20.10.16 14:10, Nelson Lameiras wrote:
> > Hello,
> > 
> > Any update on this issue?
> > Can I do anything to help?
> > 
> > cordialement, regards, 
> > Nelson LAMEIRAS 
> > 
> > Lyra Network 
> > Service Projets et Processus 
> > Tel : +33 (0) 5 32 09 09 70 
> > 109 rue de l’innovation 
> > 31670 Labège - France 
> > www.lyra-network.com
> > 
> > ----- Original Message -----
> > From: "Michal Skrivanek" <michal.skrivanek at redhat.com>
> > To: "Tomas Golembiovsky" <tgolembi at redhat.com>
> > Cc: "Nelson Lameiras" <nelson.lameiras at lyra-network.com>, users at ovirt.org
> > Sent: Tuesday, October 11, 2016 4:11:36 PM
> > Subject: Re: [ovirt-users] can't import vm from KVM host
> > 
> > > On 11 Oct 2016, at 16:01, Tomáš Golembiovský <tgolembi at redhat.com> wrote:
> > > 
> > > Hi,
> > > 
> > > On Mon, 10 Oct 2016 12:21:47 +0200 (CEST)
> > > Nelson Lameiras <nelson.lameiras at lyra-network.com> wrote:
> > > 
> > >> hello michal,
> > >> 
> > >> Yes, both paths are correct and working on the source host since the VM starts and works correctly on source host. Nevertheless I have checked with fdisk and mount to assure that these devices are indeed reachable.
> > >> 
> > >> Please understand that my setup is the following:
> > >> source host : KVM (hosting the VM to migrate)
> > >> target host : oVirt 4.0.4 (patched)
> > >> 
> > >> The error "no storage vol with matching path '/dev/sdc'" is returned by the target host script /usr/lib/python2.7/site-packages/vdsm/v2v.py.
> > >> This almost seems normal since the device /dev/sdc is not mounted on target host (where script is running) but only on source host.
> > >> 
> > >> Can this be the source of the error ?
> > > 
> > > To me this looks more like a bug in the engine, not in VDSM.
> > > 
> > > The exception in the VDSM log is OK and it is not fatal. It is actually
> > > to be expected for block devices. VDSM is trying to find out the size of
> > > the disk, but the way it does that is tailored for volumes in storage
> > > pool. But the block device is not a volume in a storage pool and that is
> > > why libvirt complains. We can potentially skip the check for block
> > > devices to avoid error in the log, or provide some other logic of
> > > getting the size.
> > > 
> > > 
> > > But the real problem is this error in engine.log:
> > > 
> > >> 2016-09-29 14:13:48,637 ERROR [org.ovirt.engine.core.bll.GetVmsFromExternalProviderQuery] (default task-30) [] Exception: org.ovirt.engine.core.common.errors.EngineException: EngineException: java.lang.NumberFormatException: null (Failed with error ENGINE and code 5001)
> > > 
> > > We don't send the size (or rather send null value) for the disk and
> > > engine does not like that. I assume engine does not really consider all
> > > the optional VM properties as optional.
> > 
> > is it really optional in schema?
> > even if it is, indeed it’s wrong:)
> > 
> > the best would be to handle it on v2v.py side first to avoid libvirt err, but i don’t know if we can get that info via libvirt in any other way:/
> > 
> > > 
> > > 
> > > 
> > >    Tomas
> > > 
> > > -- 
> > > Tomáš Golembiovský <tgolembi at redhat.com>
> > _______________________________________________
> > Users mailing list
> > Users at ovirt.org
> > http://lists.ovirt.org/mailman/listinfo/users