SATA RAID disk fail detection

[Henrik Østergaard Madsen <Henrik@Ostergaard.net>]

I have been running a few RAID 1 SATA-based servers using kernel 2.6.x
(Debian stable and testing), Linux kernel software RAID and a SATA
controller on an ASUS motherboard.

I have on several occasions have problems with a cable (and a single disk
crash), causing the box to become irresponsive, as it used 99.999% CPU to
fail accessing the disk controller (the error text was unfortunately of
cause not in the log). I did not have unusual problems replacing disks and
cables and recovering the array.

But I kind of hoped that the configuration would have failed the bad disk
and continued to run on the good one. This did not happen, as the bad
cable appearently took down the controller, thereby loosing the other disk
as well.

Is there a way to avoid this? Would it e.g. be solved by bying another
SATA controller and then placing one of the two disks (or both) on this?
Are some of them more recommendable than others, and will I still be able
to boot from both disks?

Regards

Henrik