[et-mgmt-tools] Re: strange virt-install/koan issue

Michael DeHaan mdehaan at redhat.com
Wed Jul 16 19:34:31 UTC 2008


Brenton Leanhardt wrote:
> +++ Brenton Leanhardt [16/07/08 11:13 -0400]:
>> +++ Michael DeHaan [16/07/08 11:05 -0400]:
> <snip>
>>> Brenton Leanhardt wrote:
>>>
>>> Can you share your ~/.koan/*.log and also the versions of the 
>>> associated packages?
>>>
>>> This can possibly help identify what koan should be doing with the 
>>> virt-install APIs, but maybe isn't, in your case.
>>>
>>> --Michael
>>
>> koan-1.0.1-1.fc9.noarch
>> kvm-65-7.fc9.i386
>> python-virtinst-0.300.3-7.fc9.noarch
>> kernel-2.6.25.10-86.fc9.i686
>>
>> I've attached the logs though I think I finally found something
>> interesting. I copied the temporay kernel, initrd, and xml used 
>> during the koan
>> process. I then made one modification to the xml, removing the
>> "arch='i386'" attribute from the os type. I noticed the xml created
>> by the virt-install cli didn't have that so I took a shot in the dark.
>> Seconds later I was actually able to run 'virsh start mydomain.xml'
>> and have it work. I suspect that if we can get koan to not set that
>> attribute it will solve my problem.
> <snip>
>
> More interesting advancements in tracking down this issue:
>
> 1) Removing "arch='i386'" allows for kvm guests to be
> provisioned
> 2) Changing "arch='i386'" to "arch='i686'" also allows kvm guests to
> be provisioned on my hardware
>
> After digging through all the libraries involved in the koan process I
> couldn't determine a _good_ way to implement #1. As for #2 I'm
> starting to wonder if my problem isn't related to how 'cobbler import'
> works. 'i686' is not a valid architecture for cobbler distros while both
> 'i386' and 'x86' are.
> Here's something interesting from koan/qcreate.py:
>
> if arch is not None and arch.lower() == "x86":
> arch = "i686"
>
> Unfortuntely it seems like 'x86' is never returned from Cobbler for a
> profile's arch, only 'i386', therefore the arch never get's set to
> 'i686' in my case. As mentioned in #2, if "i686" could somehow be
> set, I think koan would work for me.
> Questions:
> 1) Why does a kvm guest with a 'i386' arch not work on my hardware yet
> 'i686' does?
> 2) Should "cobbler import" import 'i386' distros as 'x86'?
> 3) Should cobblerd return 'x86' whenever the arch is 'x86' or 'i386'?

Questions aside, basically it's about making sure we always send i686 
and not the internal arch
storage. The reasons for this internal storage are really not that 
important, so I'll take care of this.
Thanks!

--Michael.




More information about the et-mgmt-tools mailing list