[Nottingham] Big biiig problem regarding IDE-RAID

Graeme Fowler nottingham at mailman.lug.org.uk
Tue Jul 15 14:25:01 2003

On Tue, 15 Jul 2003, Rob Andrews wrote:
> You may be aware that the majority of the low-end IDE RAID chipsets
> (HighPoint, Promise, etc) are not real RAID chipsets at all.

Until very recently I wasn't at all, but I am now :)

> But as far as I'd be concerned... I'd have used Linux's own software RAID
> instead of a vendor's binary-only software RAID module.

"Using"? Just testing.

This is in an effort to stop using stupidly expensive (read: 3000 quid plus) 
systems where sub-1000 quid systems will do. Without giving too much away to 
the LUG as a whole, with disk sizes growing and prices falling, large capacity 
IDE drives have just become a far more attractive proposition.

And for balance, I've seen a higher percentage of SCSI drives fail in the last
three years than IDE. They've probably been in more highly loaded,
disk-intensive systems, but still. Losing 36GB of SCSI RAID array[0] at 4am
and affecting several thousand people makes me want to investigate almost
*any* method of improving both costs and resilience!

[0] yes, I did say array. Not disk, array. It wasn't pleasant.

So for the time being this is a system under eval. If we find any more 
funnies, I'll let you know... especially as we're just upgrading the kernel :)