Gaby, please be aware of this thread:
http://marc.info/?l=openbsd-misc&m=120349790515559
AFAIK the LSI/bioctl hot spare issue still exists. If you have created
your hot spare using bioctl -H, it *will* look like a hot spare, but it
probably will *not* act like one.
You might want to give it a test to confirm, and then use the workaround
described in that thread to correctly recreate the hot spare (it will
require one reboot).
Matthew
On Wed, 18 Jun 2008, Gaby Vanhegan wrote:
We had a drive failure on a RAID5 (LSI MegaRaid SATA 150-4) volume in our
server (OpenBSD 4.1/x86). The hot spare kicked in and the volume rebuilt
fine after a successful fsck in single user mode. We put in a new drive as
the new hot spare:
# bioctl -Div ami0
bioctl: cookie = 0xd2a23c10
bio_inq
bio_inq { 0xd2a23c10, ami0, 2, 4 }
Volume Status Size Device
ami0 0 Online 501991079936 sd0 RAID5
0 Online 250995539968 0:0.0 noencl <Maxtor 6V250F0
VA11>
'V594LE9G'
1 Online 250995539968 0:1.0 noencl <Maxtor 6V250F0
VA11>
'V5075JVG'
2 Online 250995539968 0:3.0 noencl <Maxtor 6V250F0
VA11>
'V5064EEG'
ami0 1 Hot spare 250053918720 0:2.0 noencl <WDC
WD2500AAKS-00VSA01.0>
' WD-WMART1158126'
#
The thing is the hot spare is fractionally smaller than the other drives,
which is what happens when you go into a shop and ask for a 250G drive.
What's going to happen if another drive fails and the RAID array tries to
rebuild onto the slightly smaller hot spare? Will it explode or just error
out? Do we need to go back and put a slightly larger drive in?
I know this isn't the ideal place to ask the question but I figure we can't
be the only people running LSI cards under OpenBSD. So far I can't find any
good references on the 'net but my logic and intuition tells me that the
drive needs to be bigger...
G.
--
Being drunk is feeling sophisticated without being able to say it.
http://www.playr.co.uk/
