Thanks Andy for this information. I setup a Fedora 16 libvirt host and it is 
now properly converting the vmdk files to the qcow2 format. Also making the 
change from the Jira issue fixed the problem of old VM's not being deleted.

In comparing provisioning times between libvirt and ESXi, I've observed it's 
taking about twice as long to provision on the libvirt host as an ESXi host for 
Windows images, not counting the time for initially converting the vmdk files. 
The libvirt host is using the same NAS datastores that the ESXi host is using. 
Has anyone else seen this kind of difference?

Also, what are the implications for images created on a libvirt host being able 
to run on an ESXi host? Does the imaging convert to vmdks for the repository?

Thanks,
Mike

Mike Waldron
Systems Specialist
ITS - Research Computing Center
University of North Carolina at Chapel Hill

________________________________
From: Andy Kurth [[email protected]]
Sent: Monday, November 12, 2012 9:34 AM
To: [email protected]
Subject: Re: libvirt qemu-img

As Aaron mentioned, we are running Fedora 16 because of the problems converting 
images.  You need one of the latest versions of qemu-img in order for it to be 
able to convert multi-file vmdk's.  When the module was being developed last 
spring, the version included in CentOS and RHEL failed to properly convert the 
images.  Fedora 16 does include a working version.  Since compiling libvirt/KVM 
can be quite a chore, we just run Fedora 16.

Regarding the VMs not being deleted, I think you encountered this issue:
https://issues.apache.org/jira/browse/VCL-641

The underlying problem resides in get_domain_info.  Please try making the 
change explained in the Jira issue and reverting the change you made to 
delete_existing_domains.  This will be fixed in 2.3.1.

-Andy

On Mon, Nov 12, 2012 at 8:53 AM, Yannick Charbonneau 
<[email protected]<mailto:[email protected]>> wrote:
Thanks,

I ran into another issue; vms were NOT being destroyed before new ones were 
reloaded, causing vms to pileup on the host.

If found problems on our setup in libvirt.pm<http://libvirt.pm>;

Issue #1;

The statement "next if ($domain_name !~ /^$computer_name:/);" in sub 
delete_existing_domains didn't work for me.  I had to replace it with "if 
($domain_name =~ $computer_name) {" ..."}" (this might have risks involved, if 
the image has $domain_name in it name).
Issue #2

Then I had to ADD a bunch of "$domain_name =~ s/^\s+|\s+$//g;" where virsh was 
being called.  This was required because $domain_name had leading spaces which 
would make virsh fail.

Again, I think most of those issues are caused by us using older version of 
libvirt and qemu/KVM.

Existing vms are now destroyed before loading new ones in their slot.

Cheers

Yanik



________________________________________
To: [email protected]<mailto:[email protected]>; 
[email protected]<mailto:[email protected]>
Subject: RE: libvirt qemu-img

Thanks,

I ran into another issue; vms were NOT being destroyed before new ones were 
reloaded, causing vms to pileup on the host.

If found roblems on my setup in libvirt.pm<http://libvirt.pm>;

Issue #1;

The statement

-----Original Message-----
From: Aaron Peeler [mailto:[email protected]<mailto:[email protected]>]
Sent: Monday, November 12, 2012 7:21 AM
To: [email protected]<mailto:[email protected]>
Subject: Re: libvirt qemu-img

We are using Fedora 16.

Aaron

On Fri, Nov 9, 2012 at 3:39 PM, Waldron, Michael H 
<[email protected]<mailto:[email protected]>> wrote:
> I have run into the same issues in experimenting with libvirt on our
> VCL 2.3 test system. I would also be interested to hear how others are
> doing it. I'm also using CentOS 6.
>
>
> Mike Waldron
> Systems Specialist
> ITS - Research Computing Center
> University of North Carolina at Chapel Hill
>
> ________________________________
> From: Yannick Charbonneau [[email protected]<mailto:[email protected]>]
> Sent: Friday, November 09, 2012 2:51 PM
> To: '[email protected]<mailto:[email protected]>'
> Subject: RE: libvirt qemu-img
>
> I think I figured it out…
>
>
>
> It appears the qemu-img distributed with CentOS 6 is old and doesn’t
> support multi file .vmdks convert.
>
>
>
> I recompiled the latest version and it copied/converted the .vmdks properly.
>
>
>
> I am now facing a problem with libguestfs, NOT recognizing the OS
> inside the .qcow2.  Again, I think this is due to older release of libguestfs.
>
>
>
> What distro are you guys using for libvirt hosts?
>
>
>
> From: Yannick Charbonneau 
> [mailto:[email protected]<mailto:[email protected]>]
> Sent: Thursday, November 08, 2012 9:47 PM
> To: <[email protected]<mailto:[email protected]>>
> Subject: libvirt qemu-img
>
>
>
> Hi All,
>
>
>
> We have a vorking vcl 2.3 + esxi setup working.
>
>
>
> We are now trying out a libvirt host.  We have the host up but we are
> having issues deploying our vmare images on it.
>
>
>
> It appears be related with multiple files .vmdks disks and qemu-img convert.
> The resulting qcow2 file is way too small, and the vm fails to boot
> saying disk is NOT bootable.
>
>
>
> If I take a -flat.vmdk image and convert it, I can boot the vm
> manually in virt-manager.
>
>
>
> Anybody has seen this?
>
>
>
> Thanks
>
>
>
> Yanik



--
Aaron Peeler
Program Manager
Virtual Computing Lab
NC State University

All electronic mail messages in connection with State business which are sent 
to or received by this account are subject to the NC Public Records Law and may 
be disclosed to third parties.

Reply via email to