From mboxrd@z Thu Jan 1 00:00:00 1970 From: John Robinson Subject: Re: Samsung F1 RAID Class SATA/300 1TB drives Date: Thu, 04 Nov 2010 01:54:41 +0000 Message-ID: <4CD21261.8090701@anonymous.org.uk> References: <61.E7.19545.3D73FCC4@cdptpa-omtalb.mail.rr.com> <4CD19603.4060308@tmr.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4CD19603.4060308@tmr.com> Sender: linux-raid-owner@vger.kernel.org To: Bill Davidsen Cc: 'Linux-RAID' List-Id: linux-raid.ids On 03/11/2010 17:04, Bill Davidsen wrote: > Leslie Rhorer wrote: [...] >> I'm not entirely sure when the drives all went bad, but it was >> within a week or two of each other. > > One possible cause for this is a marginal power supply which "can't keep > up" when supporting lots of seeks and transfers on multiple drives. In > every group of drives there will be some variance for low voltage (or > noise, more likely) and the drive(s) which are sensitive appear to fail. > I say this from experience, it does happen, and going to a better power > supply will cure it. This might not be the problem, of course, but it's > worth investigating before blaming the drives. Yes, marginal PSUs can be, well, marginal, but it seems even half decent PSUs can exhibit this behaviour as well, to some extent. So having read the above I thought I'd look up some data sheets to see why this might happen. The spec for Seagate Constellation ES 1GB drives - http://www.seagate.com/staticfiles/support/disc/manuals/enterprise/Constellation%203_5%20in/100516232f.pdf - does say that while the average idle power is 5W[1], and the typical peak operating power is 7W, the maximum transition power (whatever that is) is 40W - yes, 40W - so I can well see a bundle of relatively low-power drives placing some heavy stresses on an average or even average-to-good PSU. My rule of thumb of late is to expect 7.2k drives to draw ~7W, 10k ~10W and 15k ~15W, then add a margin for safety and power-up, but having read the above noted I might double it and make the margin bigger... Cheers, John. [1] Rounded up to the nearest W.