From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrik Jonsson Subject: Re: Strange behaviour on "toy array" Date: Mon, 16 May 2005 14:54:01 -0700 Message-ID: <42891679.2090902@ucolick.org> References: <20050516184049.90DDA11416@smtp.ucolick.org> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20050516184049.90DDA11416@smtp.ucolick.org> Sender: linux-raid-owner@vger.kernel.org To: Ruth Ivimey-Cook , linux-raid@vger.kernel.org List-Id: linux-raid.ids Ruth Ivimey-Cook wrote: > >Yes, I believe this interpretation is correct. Moreover, I've seen this happen >"for real": when two drives died on my raid5 array while I was playing, I >started to see some I/O errors, but only for things that hadn't just been >accessed: recently accessed things were returned fine. As time went by, even >those disappeared. > >I must admit, it's rather disconcerting, but it is a logical result of having a >block cache. > > This makes sense, however, I would have expected /proc/mdstat or something telling me the array is DEAD. It seems "clean, degraded" is not a proper description of a raid5 without any working drives... Or would this not happen until I tried to write to it (which I haven't gotten to yet)? I must admit I don't remember seeing in the FAQ or anywhere what is supposed to happen when you lose more than one drive. I sort of expected to have the entire array go offline, but it seems it just limps along like a normal faulty drive would do? /Patrik