[lug] upgrading OS on RAID1

Lee Woodworth blug-mail at duboulder.com
Tue Mar 8 16:04:53 MST 2005


Lori Reed wrote:
> 
> I did this as a matter of routine when I was the system architect for a 
> voice mail/voice processing system. However, the system I designed used 
> an external scsi raid 1 controller, so it was mostly invisible to the
> o/s at runtime and completely invisible to the bios at boot time.
> 
> I presently have 2 mobos with on-board ide raid 1 controllers, but I've 
> never used this feature so I don't know how the bios responds at boot 
> time. Since both of these boards also support booting from drive D, I 
> presume the right thing will happen if either the primary or secondary 
> drive is absent at boot.
Depends on the chipset, but the ones I have used (Promise, HPT) do 
present the raid set as a BIOS disk. It can even be the first drive
or the only drive visible.

> 
> One thing that worked so well in my raid 1 days that I still do it today 
> is to put all hard drives in removable carriers.
> 
> IMNSHO, software raid is seldom, if ever, a good idea.
That depends on the raid hw. I ended up with corrupted file systems 
because of cheapo mobo raid controllers (HPT). One disk in a mirror 
devloped bad blocks and the controller didn't handle it correctly. I 
think successive reads to the same block came from different disks and 
had different contents but the controller sailed along as if everything 
was fine.

I think cheap controllers that can't do their own bad block
scans/relocates are a disaster waiting to happen. On the other hand they
can perform well. I was getting 22MB/s sustained transfers (a few gigs) 
out of a Promise Fastrak33 with a 2 disk stripe set (20GB, 5400RPM WD
Caviar). This was about 7 years ago.

> 
> Lori



More information about the LUG mailing list