[Nottingham] Big biiig problem regarding IDE-RAID

Robert Davies nottingham at mailman.lug.org.uk
Tue Jul 15 13:40:00 2003


On Tuesday 15 Jul 2003 12:33, Alex Herington wrote:

> Lessons I have learned (heh): Don't panic so much. Heart attacks at 21
> aren't cool.. and IDE-RAID isn't worth the hassle (with some Promise
> chipsets at least). Spend the extra dough and go SCSI.

Those IDE RAID cards, are really software driver based, so they don't give you 
anything really extra over Linux Software RAID (which at least allows you to 
spread the disks over seperate IDE controllers and avoid having your data 
being kept in a proprietary format).

Both disks developing bad sectors is something you want to look into, frankly 
I don't believe it!!  Either you have managed to snag an IDE cable and 
getting data errors, the controller and/or disks overheated, hot weather is 
the room well vented or got a/c?

I've also seen cases where partition tables have got corrupted, which caused 
lots of errors, but fortunately sfdisk(8) was able to fix up the table, to 
allow recovery of all data.

Finally there have been problem in some recent-ish kernel versions with the 
Promise drivers, I think 2.4.19 - 2.4.20 roughly were affected, some kernels 
couldn't boot them, so if you recently did a kernel upgrade, you might want 
to look into driver issues as a posibility.

On data loss with RAID, believe me a badly thought out SCSI RAID Tower system, 
can be far more lethal to your data than anything that you've experienced.  
Especially in the days when the SCSI drivers (2.2.12-ish) would ignore read 
errors because they couldn't report them properly just retry, and leading to 
writes being copied to random blocks.  The system would stop (usually) when 
it had to have the VM page in from swap.

Rob