I am wide open to suggestions (see discussion at bottom, as usual).
On 1/19/2014 5:25 AM, Gianluca Cecchi wrote:
On Sun, Jan 19, 2014 at 7:13 AM, Ted Miller wrote:
* BEAT HEAD AGAINST WALL because virt-v2v.x86_64 0.9.1-5.el6_5 from Centos
updates doesn't seem to know about .ova files. I was following the
instructions in
Red_Hat_Enterprise_Virtualization-3.3-Beta-V2V_Guide-en-US.pdf guide, but I
figured out that the v2v they are talking about has an "-i ova" option,
while the help file for the version I am using does not list ova as an
option for -i, and if I try to use it, it tells me that it is an invalid
option, and if I leave it off it goes off looking for a qemu///system to
attach to. help files for v2v say nothing at all about .ova files.
I am wondering where to find a v2v program that knows about .ova files, or
else am I going to have to import all my VMWare files to my (non-ovirt) KVM
host, and then drag them into ovirt from libvirt?
I made a bit of research about this
Strange.... I just update a CentOS 6.4 VM to latest 6.5 and see that
there (also matching RHEL 6.5 I think) there is indeed as you wrote:
virt-v2v-0.9.1-5.el6_5.x86_64
And it seems ova is missing as an option...
Instead on a Fedora 19 system with
virt-v2v-0.9.0-3.fc19.x86_64
I have it....
So for any reason was it removed in newer packages?
It seems also strange to see a Fedora package (even if 19 and not 20)
older than a RH EL 6 one ...
RHEL 6 version bumped this way skipping 0.9.0:
* Wed Jun 12 2013 Matthew Booth <[email protected]> - 0.9.1-1
- Rebase to new upstream release
* Mon Oct 22 2012 Matthew Booth <[email protected]> - 0.8.9-2
while fedora 19 has been currently stopped at
* Wed Jul 03 2013 Richard W.M. Jones <[email protected]> - 0.9.0-3
- Default to using the appliance backend, since in Fedora >= 18 the
libvirt backend doesn't support the 'iface' parameter which virt-v2v
requires.
- Add BR perl(Sys::Syslog), required to run the tests.
- Remove some cruft from the spec file.
BTW in F20 we do have ova too:
virt-v2v-0.9.0-5.fc20.x86_64
and in fact it has the older version...
For RHEL 6 I remained here:
http://lists.ovirt.org/pipermail/users/2013-May/014457.html
ANd no particular virt-v2v package in rhev source repo
http://ftp.redhat.com/redhat/linux/enterprise/6Server/en/RHEV/SRPMS/
For sure the rhev 3.3 beta guide is incorrect at the moment
https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Virtualization/3.2/html-single/V2V_Guide/index.html#chap-V2V_Guide-Installing_virt_v2v
because it says
"
virt-v2v is available on Red Hat Network (RHN) in the Red Hat
Enterprise Linux Server (v.6 for 64-bit x86_64) or Red Hat Enterprise
Linux Workstation (v.6 for x86_64) channel. Ensure the system is
subscribed to the appropriate channel before installing virt-v2v.
"
and some lines below
"
7.1. virt-v2v Parameters
The following parameters can be used with virt-v2v:
-i input Specifies the input method to obtain the guest for
conversion. The default is libvirt. Supported options are:
libvirt
Guest argument is the name of a libvirt domain.
libvirtxml
Guest argument is the path to an XML file containing a libvirt domain.
ova
"
Any light to shed on this?
Thanks
Gianluca
I think I have some light, but you'd better get out your rose-colored
glasses, because that is the only way the light will look good. ;(
I spun up a Fedora 20 64-bit VM (in my brand new oVirt environment) to take
advantage of your wonderful discovery. I did a minimal install, then "yum
upgrade", then "yum install virt-v2v" brought in 489MB of dependencies! This
is what I found (not necessarily in the order I found them).
1. virt-v2v has a missing dependency: perl-Archive-Tar
yum install perl-Archive-Tar
2. virt-v2v would error out fairly early in the conversion process:
Error extracting archive '/media/VMold/vmware.dud/Fedora13A.ova':
/usr/bin/tar: Fedora13A-disk1.vmdk: Wrote only 6144 of 10240 bytes
that error went away when I increased VM memory from 1G to 4G (see below).
3. The *.ova files produced by vmware-vdiskmanager build 835872 (downloaded
yesterday as part of VDDK 5.0) give the following error messages when running
through virt-v2v:
Use of uninitialized value $file in hash element at
/usr/share/perl5/vendor_perl/Sys/VirtConvert/Connection/VMwareOVASource.pm
line
261, <$manifest> line 2.
Reading from filehandle failed at
/usr/share/perl5/vendor_perl/Sys/VirtConvert/Connection/VMwareOVASource.pm
line
271.
Though I am not a Perl programmer, I took a look at the code, stuck in some
debugging "print" statements, and came to this conclusion:
The *.ova files produced by my version of vmware-vdiskmanager contain a
"blank" line at the end of the manifest with about 62 spaces in it
(nothing else). "sub _verify_manifest" in the file "VMwareOVASource.pm"
it throws the error.
Since the line is supposed to have and "=" in it, I added "if ($line =~
/=/){ .... } around lines 260-262. It works, though it may be very bad perl.
4. The perl code apparently attempts to extract the ENTIRE virtual disk into
memory. After implementing #3, I got errors:
line 109 Error extracting archive
'/media/VMold/vmware.dud/Fedora13A.ova': /usr/bin/tar:
Fedora13A-disk1.vmdk: Wrote only 2048 of 10240 bytes
/usr/bin/tar: Exiting with failure status due to previous errors
at
/usr/share/perl5/vendor_perl/Sys/VirtConvert/Connection/VMwareOVASource.pm
line
114.
Out of memory!
(in cleanup) open3: fork failed: Cannot allocate memory at
/usr/share/perl5/vendor_perl/Sys/VirtConvert/ExecHelper.pm line 75.
I increased the VM memory from 1G to 4G. Processing a 6-8G virtual disk,
I ran "top" and watched the physical memory go from 3+G to 128K, then
watched the 1G of swap memory disappear, then got the error message above.
5. I chose one of my smallest .ova files (less than 1G), and it fit into
memory, so got past that error. Then I got to the final insult:
virt-v2v: RHEV cannot handle volumes of format vmdk
At that I threw in the towel.
I thought about filing on Bugzilla, but I don't know whether to file 1 bug or
5 or ....
To answer your question Gianluca, yes, I think I know why they pulled out the
*.ova support -- it doesn't work yet.
I checked, but VMWare still seems to be using the *.vmdk naming convention
for their disks, so I don't think I can find another name to make virt-v2v
happy. For now it looks like I am stuck, unless I bounce everything over to
my KVM+virt-manager host, and then import it from there.
That makes two days (at well over 8 hours each) that I have spent trying to
import my VMs, and so far I have nothing to show for my work.
Ted Miller
Elkhart, IN
_______________________________________________
Users mailing list
[email protected]
http://lists.ovirt.org/mailman/listinfo/users
