Big thanks to all those who responded, I am about to replace a member
of a raid set and was looking for some re-assurance I was 'doing the
right thing'.
----- Original request -----
Just a quick sanity check...
I have an AS800 with a single channel raid controller with 4 disks
attached. Disks are configured as a single member JBOD, and a 3 member
RAID5 set. One of the RAID5 members has started clocking lots of 'misc'
errors and a few 'soft' errors, it's on maintenance so I'm going to
replace it.
Current thinking is I can just replace the faulty drive, bring the system
back up, mark the new disk as optimal and let the system rebuild it
automatically, is this correct?
----- Response 1 -----
> From: "O'Brien, Pat" <pobrien at mitidata.com>
>
> It is nice to dream. It really is a relationship issue where the disks
are
> in target bus relationship, and How they are defined within the
backplane
> raid controller. If the raid set is the first defined object, you need
to
> delete the others in higher bus order to replace and re-define the
deleted
> items. Those controllers get clunky for this type of service.
In my case the raid set is the last defined object, and I'm doing a 1
for 1 replace, so I think I don't need to worry...
----- Response 2 -----
> From: alan at nabeth.cxo.cpqcorp.net
>
> You may be able to replace the disk without having to
> shutdown the system. It depends on the subsystem and
> the drive enclosure. For our array controllers, SWCC
> may allow quieting the bus for the moment it takes to
> swap the disk. While StorageWorks SBBs and enclosures
> are design to support hot swap, there is a low risk or
> perturbing other I/O on the SCSI bus and it is wise to
> quiet the bus. The controller documentation may also
> how best to swap a disk.
>
> I'm not sure that making a new disk "optimal" is exactly
> the right way to get it start rebuilding the data from
> the others, but that should be documented. I worry about
> the "mark optimal" bit because we sold a subsystem a long
> time ago where this meant "trust me, the data on this is
> correct", instead "rebuild the content of this disk from
> the data on the others".
It's ok to take the system down, so I'll do that anyway, but thanks
for the note. Following later messages (below) I'll mark the disk as
hot spare and see how we go from there.
----- Response 3 -----
> From: Ron Bramblett <bramblet at fuller.com>
>
> That sounds right to me but you might have to use the Raid configuration
> utility to mark the disk as bad then boot again to the raid utility and
> have it rebuild it.
>
> Some systems have auto rebuild but I am not sure. It depends on your
raid
> controller how it is handled.
Ok, I can see what happens, shouldn't be a problem to kick it off by
hand it it doesn't 'do the right thing' itself.
----- Response 4 -----
> From: Jim Kurtenbach <KurtenbachJ at crlcorp.com>
>
> I always add the new disk in and configure it as a hotspare. The raid
> controller should pick it up and rebuild it.
This seems almost obvious in retrospect, I'll go this route and all
will be groovy I'm sure...
Thanks again,
Simon
--
Simon Greaves Voice: +679 212114
Systems & Networks Fax: +679 304089
ITS, USP, Suva Email: Simon.Greaves_at_usp.ac.fj
Fiji
Received on Mon Jan 28 2002 - 21:54:23 NZDT