Re: [RE]Intel SATA RAID

James Olson wrote:
> That solution of stopping and starting dmraid on the same line was 
> only tested for those people who had previously installed FC5 to their 
> striped dmraid array, and were unable to get it to work with a new 
> kernel 2.6.17 and above update. If you created your partitions 
> manually and manually copied the data over then there might be other 
> problems to deal with. I suggest trying to install FC5 to the array 
> with anaconda, the fedora core installer, first.
>
>
Okay, been there, done that. Same problem. New twist.

I nailed the container, recreated it (new name now), then installed a 
minimal FC5 and used anaconda to create a custom layout.
Rebooted,  updated dmraid to rc11, installed the new 2.6.18-1 kernel, 
rebooted, nada.
Had a peak in the init file for the two (base/new kernel) initrd images 
and found the following missing from the new one.

rmparts sdb
rmparts sda
dm create isw_cdgbeijgjh_iaa 0 1250275846 striped 2 256 8:0 8:16 0
dm partadd isw_cdgbeijgjh_iaa

ah ha! I thought, that's what I need. The container setup is missing 
from the new kernel, why did the mkinitrd miss that? anyway, copied 
those lines into the new kernel init, repacked the init rd and rebooted.
Then I get.  device-mapper: table: 253:0 striped: Target length not 
divisble by chunk size

Okay, so for some reason the new kernel (which runs device-mapper 4.7) 
doesn't like my RAID partitioning either. *sigh*

So I've subscribed to the dm-devel list and asked for help there. I 
searched their archives and found a couple of references, these usually 
ended with people getting patches and the like.

Hopefully they can shed some light on this. But I can rule out
- dmraid level, now rc11
- parttition creation, done in anaconda

On the upside, I did identify the missing lines from the initrd and 
managed to add them in quickly, just a pity that didn't fix it. ;-(

Thanks for your time and patience. Hadders