We had endless problems with SCSI cables on our older arrays. They work
lose due to thermal cycling, the contacts become tarnished (even when
connected tight) and generally, the cables 'wear'. When running at full
speed, scsi cables are the weakest link - the better the cable (read
more expensive), the more likely it is to stand up to higher
performance.
Your first action should be to stop the system, unplug all the scsi
cables, clean the contacts, and reseat them. If that doesn't work,
clean the contacts of and reseat the raid controller in the array too -
it will suffer the same problems of dodgy contacts.
Martin.
--
Martin Bly
RAL Tier1 Fabric Team
> -----Original Message-----
> From: Testbed Support for GridPP member institutes
> [mailto:[log in to unmask]] On Behalf Of Kostas Georgiou
> Sent: 06 August 2008 11:49
> To: [log in to unmask]
> Subject: Re: SCSI mounted RAID problems
>
> On Wed, Aug 06, 2008 at 11:32:46AM +0100, Matt Doidge wrote:
>
> > Our pool nodes are CV provided Supermicros mounting a pair of 6TB
> > Infortrend Eonstor A16U-G1A3 via U-320 scsi cables via a
> PCI Adaptec
> > 39320A-R U320 SCSI Card.
>
> From what I remember the Adaptec firmware has issues with the
> Infortrend
> boxes, there is an faq entry about it in the Infortrend web site. It's
> been three or more years since I last looked at it so a newer firmware
> might be available but at the time the workaround was to
> lower the speed
> of the scsi bus.
>
> Cheers,
> Kostas
>
|