On Tue, 11 Apr 2000, Ed Carp wrote:
> =?iso-8859-1?Q?Jakob_=D8stergaard?= (jakob@ostenfeld.dk) writes:
>
> > I'm just setting up such a box. With two IDE disks, and software RAID-1. I can
> > remove any disk and it will still boot as if nothing happened. When bad blocks
> > are detected, the RAID layer kicks the disk/partition in question, and the
> > system keeps going. I've seen this work in Real Life before. Sure you need to
> > plan 15 minutes of downtime at some point in the future to replace a failed
> > disk, but you can actually _plan_ it.
>
> I don't that sort of luxury. The hardware's already out in the field. I have
> to make do with what I have. If I go back and tell my client that, yes, if
> we have a single-bit failure on a disk sector, our only alternative is to go
> to two drives on the card and do software RAID, I'll be fired (and rightly so)
> for incompetence.
>
> Such "perfect world solutions" simply don't exist sometimes in the real world.
RAID can still help you out. Build a RAID-1 or -5 on N partitions on the
_same_ drive. Performance will suck, but once bad sectors start appearing,
only the partition in question will be kicked. With a N partition RAID-1, you
can still survive a N-1 partition failure without losing a _single_ bit of
information. And without the need of any patching of ext2fs at all. It's all
in the kernel, today. (For 2.2. you should use the RAID patch from
http://people.redhat.com/mingo though)
Buy a 20 Gig drive, make it a 10 partition RAID-1, and you have 2 GB of fairly
reliable storage that can survive bad sectors but not a head crash.
Bad sectors are often localized physically on the platter, and therefore also
localized on the partitions.
--
................................................................
: jakob@ostenfeld.dtu.dk : And I see the elder races, :
:.........................: putrid forms of man :
: Jakob Østergaard : See him rise and claim the earth, :
: OZ9ABN : his downfall is at hand. :
:.........................:............{Konkhra}...............:
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Apr 15 2000 - 21:00:19 EST