Re: [Users] Importing from KVM

On 04/05/2012 03:22 PM, Shahar Havivi wrote: >On 05.04.12 12:47, Alex Drahon wrote: >>OK, I went back to the original OVF file (just had to remove the Ethernet >>adapter Item so it ovirt would accept it). >> >>I'm now back to my original error, integrity violation on pk_vm_device... >if you are adding a device I assume that you need to generate a new GUID for >it (or it will use a null one 0000:0000:000... and you will get primary key >violation on the second device). we can't require this, as it will break backward compatibility for importing OVFs. so this should be a bug.

>if its not the case so maybe Eli can help, >Eli do you know why Alex is getting the pk_vm_device? > >in any way can you post the OVF? >> >>On Thu, Apr 5, 2012 at 12:12 PM, Shahar Havivi&lt;shaharh(a)redhat.com&gt; wrote: >> >>>On 05.04.12 13:56, Itamar Heim wrote: >>>>On 04/05/2012 01:50 PM, Alex Drahon wrote: >>>>>I tried with the minimal OVF file, which means no UUID since it's >>>>>defined (I think) in >>>>><Section ovf:id="..." ovf:required="false" >>>>>xsi:type="ovf:OperatingSystemSection_Type"> >>>> >>>>shahar - uuid is not a must? >>>>if not and engine is supposed to create it, sounds like a bug? >>>engine doesn't create that, >>>it expect the UUID to be present. >>>> >>>>> >>>>>Of course it fails if ovirt-engine doesn't generate one. >>>>> >>>>>On Thu, Apr 5, 2012 at 11:44 AM, Itamar Heim&lt;iheim(a)redhat.com >>>>><mailto:iheim@redhat.com>> wrote: >>>>> >>>>> On 04/05/2012 01:42 PM, Alex Drahon wrote: >>>>> >>>>> Here's what I found in engine.log >>>>> >>>>> >>>>> are you sure you are not trying to add a VM with an existing UUID >>>>> (or did not pass a UUID)? >>>>> >>>>> >>>>> On Thu, Apr 5, 2012 at 11:32 AM, Itamar Heim&lt;iheim(a)redhat.com >>>>> <mailto:iheim@redhat.com> >>>>> <mailto:iheim@redhat.com<mailto:iheim@redhat.com>>> wrote: >>>>> >>>>> On 04/05/2012 01:16 PM, Alex Drahon wrote: >>>>> >>>>> Thanks for your help. >>>>> >>>>> I tried the import with a minimal OVF file as described >>>by >>>>> Shahar and it >>>>> worked, but failed on activate. >>>>> >>>>> This time I have a constraint violation on pk_vm_static, >>>>> which means >>>>> that ovirt-engine doesn't create a vm_guid if it isn't >>>>> here. I guess >>>>> it's the same for a lot of parameters. >>>>> >>>>> >>>>> that's sounds like a new bug. >>>>> please send stack trace (which will also help telling where >>>>> to look). >>>>> >>>>> >>>>> >>>>> To move forward, I'm going to create "shadow" VMs on >>>>> ovirt-engine, >>>>> export them and import my "real" libvirt/KVM VMs with >>>the >>>>> exported OVF >>>>> files. Makes sense? >>>>> >>>>> Where should I look in the source tree if I want to >>>(try to) >>>>> actually >>>>> help with the restore code for imported VMs? >>>>> >>>>> Alex >>>>> >>>>> On Thu, Apr 5, 2012 at 5:59 AM, Shahar Havivi >>>>> <shaharh@redhat.com<mailto:shaharh@redhat.com> >>>>> <mailto:shaharh@redhat.com<mailto:shaharh@redhat.com>> >>>>> <mailto:shaharh@redhat.com<mailto:shaharh@redhat.com> >>>>> <mailto:shaharh@redhat.com<mailto:shaharh@redhat.com>>>> >>>wrote: >>>>> >>>>> On 05.04.12 07:35, Itamar Heim wrote: >>>>> > On 04/04/2012 01:13 PM, Alex Drahon wrote: >>>>> > >Hello all, >>>>> > > >>>>> > >I'm trying to import VMs from a Centos5/KVM host to ovirt >>>using >>>>> > >virt-v2v. I used this command: >>>>> > >virt-v2v -ic >>>>> qemu+ssh://root@virthost1.____mydomain.local/system -o >>>>> rhev -os >>>>> > >virthost3.mydomain.local:/____exports/exp --network >>>ovirtmgmt >>>>> >>>>> > >test.mydomain.local >>>>> > > >>>>> > >It works, but the resulting OVF file cannot be read by >>>>> ovirt-engine >>>>> > >("Failed to read VM '[Unknown name]' OVF, it may be >>>corrupted"). >>>>> > > >>>>> > >I had to edit the OVF file, removing the NIC Item worked, >>>but >>>>> restoring >>>>> > >the imported VM fails. Looking at logs, I found a PostgreSQL >>>>> constraint >>>>> > >violation on an insert (sorry it's in French) >>>>> > >"Caused by: org.postgresql.util.____PSQLException: ERREUR: >>>la >>>>> >>>>> valeur >>>>> d une >>>>> > >clé dupliquée rompt la contrainte unique « pk_vm_device » >>>>> > > Where: instruction SQL « INSERT INTO vm_device( >>>device_id, >>>>> vm_id , >>>>> > >device , type, address , boot_order , spec_params, >>>is_managed, >>>>> > >is_plugged, is_readonly) VALUES( $1 , $2 , $3 , $4 >>> , $5 >>>>> , $6 >>>>> > >, $7 , $8 , $9 , $10 ) »" >>>>> > > >>>>> > >Looks like there's a primary key violation (duplicate) on >>>>> insert. >>>>> > >>>>> > cc-ing mbooth from v2v to have a look from ovf generation, >>>>> but there >>>>> > could be some bug in engine as well as the vm_device code was >>>>> added >>>>> > recently. >>>>> > >>>>> > > >>>>> > >I have two questions: >>>>> > > - what's the minimal OVF file that can successfully import >>>>> a VM? >>>>> you will need the header: the<Reference> tag the >>>>> <Content> >>>>> tag with the >>>>> <Section> items:<System> and<item>s cpu and memory, all other >>>>> devices can be >>>>> omit. >>>>> > >>>>> > shahar? >>>>> > >>>>> > > - are 'default devices' created if they are not in the OVF >>>>> file? >>>>> > >(would explain PK violations) >>>>> > >>>>> > eli? >>>>> > >>>>> > > >>>>> > >oVirt is very impressive and stable, deploying was a >>>breeze, but >>>>> it"s >>>>> > >difficult to understand why it is so hard to move VMs from a >>>>> libvirt/KVM >>>>> > >host (which is basically the same architecture). Am I >>>missing >>>>> something? >>>>> > >>>>> > >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>> >>>