Ovirt devel - Jun 2008 - Managed node identification working...

I've finished up the code for making a managed node identify itself to 
the oVirt server and for retrieving the keytab. However, I've hit two 
stumbling blocks and am not sure how to proceed:

1. The UUID returned is currently defaulting to the managed node's 
hostname. I'll plead ignorance and say I'm not sure how to retrieving 
the UUID for the node. I see where, with a domain reference, I can grab 
the UUID string via the libvirt APIs, but am not sure which domain to use.

2. Putting the ovirt-managed-node.rpm into the pungi repo's not happened 
for me. So, for now, the common-pkgs.ks does not include 
ovirt-managed-node and I've had to manually push the RPM onto the 
managed nodes. But, once there, they boot up and identify themselves 
just fine with the developer node.

I would appreciate a little assistance on these two pieces. Thanks. :)

-- 
Darryl L. Pierce, Sr. Software Engineer
Red Hat, Inc. - http://www.redhat.com/
oVirt - Virtual Machine Management - http://www.ovirt.org/
"What do you care what other people think, Mr. Feynman?"
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dpierce.vcf
Type: text/x-vcard
Size: 319 bytes
Desc: not available
URL:
<http://listman.redhat.com/archives/ovirt-devel/attachments/20080618/7cb3de86/attachment.vcf>

Darryl L. Pierce wrote:
> I've finished up the code for making a managed node identify itself to 
> the oVirt server and for retrieving the keytab. However, I've hit two 
> stumbling blocks and am not sure how to proceed:
> 
> 1. The UUID returned is currently defaulting to the managed node's 
> hostname. I'll plead ignorance and say I'm not sure how to
retrieving
> the UUID for the node. I see where, with a domain reference, I can grab 
> the UUID string via the libvirt APIs, but am not sure which domain to use.
Darryl and I discussed this on IRC, for those who weren't watching.  We have
two
problems with getting the UUID:

1)  In the "fake" managed node case, KVM currently assigns the same
UUID to
every single guest.  So this obviously doesn't work as any sort of unique
identifier.  danpb says there is a patch in the works for upstream Qemu to get a
-uuid parameter; once we have that, and some plumbing of that information
through to the DMI tables, we should be able to use dmidecode (either directly
or through HAL) to get the UUID of the machine.

2)  Real, physical hardware does not always include a UUID.  It's just a
matter
of whether the BIOS writer decided to do it.  Most of the big-box vendors do
tend to have a UUID, but not always, and with whitebox type hardware it is a
crapshoot.  We need to have some fallback plan if we can't find a UUID in
DMI on
real hardware.

For now, I advised Darryl to leave the UUID as hostname, until we solve these
two issues.  By virtue of having to have static hostnames for kerberos, the
"uniqueness" requirement is actually filled by hostname.

Chris Lalancette