R James skrev 5.3.2013 19:53:
On Tue, Mar 5, 2013 at 11:11 AM, Colin Guthrie <mag...@colin.guthr.ie
<mailto:mag...@colin.guthr.ie>> wrote:
Anything that relies on ordering is just broken by design. We need to
handle things gracefully regardless of the order they are detected. This
is why UUIDs are the defacto method for filesystem identification these
days and why in mga4 we'll likely switch to a consistent naming scheme
for networking devices too.
Yes, I understand that cryptic-looking UUIDs are the defacto identifiers
but when mdadm reports that /dev/sdf1 has failed in a parity RAID setup,
it will be good if a mere mortal can know which drive to replace. :o)
If you follow raid devel list you will soon learn that they dont
recommend trusting the /dev/sd* naming either as it is by
no means static... :)
depending on your hw, they may for example hange place if you happend
to have a usb disk plugged at boot and so on.
so the thing to check is for example what disk is mapped as
/dev/disk/by-id/*
where you can match on actual disc serial number and so on...
then you can be sure wich disk is failing / has failed...
--
Thomas
