AlphaCore 1.0 final release !

[Sergey Tikhonov]

Steven N. Hirsch wrote:

>On Wed, 11 May 2005, Balint Cristian wrote:
>
>  
>
>>On Wednesday 11 May 2005 15:07, Steven N. Hirsch wrote:
>>    
>>
>>>On Wed, 11 May 2005, Balint Cristian wrote:
>>>
>>>      
>>>
>>>>>I'm running into problems trying to install Alpha Core 1.0.  The box is a
>>>>>UP2000+ with an Adaptec 39160 controller.  There are (2) drives attached,
>>>>>one on each channel.  The dka device is partitioned and formatted for
>>>>>Tru64 Unix.  Dkb has (3) other Linux distributions installed and has been
>>>>>partitioned with a BSD disklabel scheme.
>>>>>
>>>>>I'm trying to install onto an unused partition on dkb.  The installer
>>>>>trips over dka, telling me that the partition table is unreadable and
>>>>>asking if I want to reinitialize it.  I answer "no".  It then tells me
>>>>>that dkb is unreadable as well! When I tell it not to initialize that
>>>>>disk, it terminates with an error (thinking that no disks are available).
>>>>>
>>>>>The question is: why can't the installer read the partition information
>>>>>on
>>>>>either drive?  Granted, it won't know about AdvFS on dka, but it still
>>>>>should have read the disklabel.  And, since the Linux distros on dkb are
>>>>>functional there's no reason it should have problems of any sort.
>>>>>          
>>>>>
>>>>Hmm interesting. I notice the issue, keep that in mind.
>>>>
>>>>        
>>>>
>>>>>Any advice as to how I proceed with installation and avoid repartitioning
>>>>>the second disk?
>>>>>          
>>>>>
>>>>As quick hack, try unplug physicaly dka down from SCSI bus !
>>>>        
>>>>
>>>Just tried that.  It still claims it cannot "read" the disk.  Sergey 
>>>recommended that I boot in rescue mode and try to mount it manually.  This 
>>>works fine, as I expected.
>>>      
>>>
>>What does it say exactly, can do Alt-F4 or some sort of Alt-F2,3,4 to see more debugs ?
>>    
>>
>
>I am now successfully installing on a fresh disk, which I permitted AC to 
>partition and format for me :-).  
>
>I notice one major difference between what Anaconda did on that disk and 
>the way the troublesome disk is setup.  Anaconda does not follow the BSD 
>convention of making the "C" slice cover the entire drive!  It has /boot 
>as A, swap as B and the root partition as C.  My drive has /boot starting 
>at cylinder 2 (to leave room for aboot), then swap as B, then C as the 
>entire drive (starting at cylinder 1, not 2), followed by all the root 
>partitions (D, E, F) from the various distros.
>
>Not being familiar with Anaconda, I wonder if it is getting confused by 
>the situation - particularly the 'C' slice?
>  
>
I think your assumptions are correct. The used anaconda is mainline one 
with very few alpha specific patches.
To fix and debug anaconda scripts is very time consuming. I didn't find 
a way (although I didn't look hard) to do it
without having a second machine.
It might not know about rules for "C slice" and simply decides that no 
free space available. :) I don't remember
was RH7.2 able to install on such patition configuration.

>The only messages which were not routine are a series of complaints about 
>Anaconda using an obsolete ioctl call for scsi generic driver.  This 
>occurs at the time it's probing for the partition information.
>
>Please note that an actual booting kernel recognizes the partition map - 
>AlphaCore 1.0 included.  If I boot in rescue mode, fdisk shows me exactly 
>what I expect to see.  This is almost certainly an Anaconda problem rather 
>than a kernel problem.
>
>Steve
>
>_______________________________________________
>axp-list mailing list
>axp-list at redhat.com
>https://www.redhat.com/mailman/listinfo/axp-list
>
>  
>
Regards,

-- 
---
Sergey Tikhonov
Solvo Ltd.
tsv at solvo.ru

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/axp-list/attachments/20050511/1c156541/attachment.htm>