Re: ahc SCSI (and raidframe) Problems

["Can Erkin Acar" <canacar_(_at_)_eee_(_dot_)_metu_(_dot_)_edu_(_dot_)_tr>]

On 9 Jan 2001, at 17:34, Peter Galbavy wrote:

> Bring the machine up, run "raidctl -P raid0" or whatever and fsck - these
> can be run in parallel, but for performance and risk-limitation, do one as a
> time.

This is the standard procedure and I do this after every crash ;-)

My question was what happens when the raid set is dirty
AND missing a drive. Would it still recover?
And how would it compute the parity with the missing drive?
I do not know the internals of raidframe but from what I know
about raid there could be problems.

Perhaps I should set up a test machine and give it a try.
 
> If the problem is a controller one, and you have more than one disk on the
> same controller, then you are on your own.

I guess I should try different SCSI controllers.
Any suggestions for a controller more stable than ahc on OpenBSD?
There was a discussion back about Adaptec vs QLogic(?)
But I need to make sure that I have a controller problem before
trying others.
 
> We use IDE w/raidframe RAID5 and each drive on its own bus. This is one of
> the things that makes me consider IDE drives "as good as" SCSI - no need to
> have multiple outstanding requests on different drives, since each
> controller does its own thing.

I am pretty much stuck with SCSI at this stage I guess
but thank you for your suggestion

Can


--=< Can Erkin Acar (canacar_(_at_)_bigfoot_(_dot_)_com) >=--