Re: preupgrade stuck

Gerry Reno wrote:
> Seth Vidal wrote:
>>
>>
>> On Thu, 23 Apr 2009, Gerry Reno wrote:
>>
>>> Seth Vidal wrote:
>>>>
>>>>
>>>> On Thu, 23 Apr 2009, Gerry Reno wrote:
>>>>
>>>>> Yes, anaconda puts the system into about the same as "rescue mode" 
>>>>> and then performs the operations there. But that is why I thought 
>>>>> that the upgrade could be completed from rescue mode because it is 
>>>>> very similar if not exactly similar to how anaconda handles the 
>>>>> upgrade.
>>>>
>>>> Anaconda does a lot of things other than that, too. Especially in 
>>>> post.
>>>>
>>>> -sv
>>> I can't leave this machine stuck like this. I have to try something 
>>> from rescue mode to get the upgrade completed. I know anaconda has 
>>> to do some post tasks but probably the biggest one is just 
>>> straightening out the bootloader and then removing the install 
>>> kernel and install.img from /upgrade. I can't see where it needs to 
>>> do any more than that.
>>>
>>
>> How about figuring out why it won't boot into anaconda for the 
>> preupgrade to begin with?
>>
>> -sv
>>
> Yes, I already know why. I was able to check the program.log from the 
> /tmp ramdisk in the ctrl-alt console and it is failing to assemble the 
> raid arrays because of a bad argument ot mdadm. It is trying to 
> assemble "/dev/md/0" instead of "/dev/md0" and the assembly fails, 
> therefore the volume groups fail, therefore the filesystems cannot 
> mount and then it reports that it cannot find the root of the existing 
> installation. I put all this in the bug: 
> https://bugzilla.redhat.com/show_bug.cgi?id=496258
>
If you look in the logs you'll see where I had to copy and modify the 
bootloader line for the Upgrade to Rawhide entry. I had to use NFS in 
order for anaconda to find the kickstart file. Otherwise anaconda hits 
the other bug that I filed where it cannot find the kickstart file even 
when you give it the proper device and path. This may also be due to it 
being located on a raid array. Anyway, using NFS got by that problem and 
then I ran into the next problem where it was trying to assemble the 
raid arrays using the wrong names.

Regards,
Gerry