From mboxrd@z Thu Jan 1 00:00:00 1970 From: Neil Brown Subject: Re: Spare drive won't spin down Date: Tue, 18 May 2010 10:50:49 +1000 Message-ID: <20100518105049.490aa652@notabene.brown> References: <922812.29139.qm@web31703.mail.mud.yahoo.com> <20100507162034.1370d645@notabene.brown> <717013.48748.qm@web31703.mail.mud.yahoo.com> <20100507194704.3ffcf17e@notabene.brown> <4BE83B75.8050709@tmr.com> <20100512065318.44e934d4@notabene.brown> <4BF186DE.5050502@redhat.com> <20100518102312.01c85274@notabene.brown> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: Sender: linux-raid-owner@vger.kernel.org To: Michael Evans Cc: Doug Ledford , Bill Davidsen , Joe Bryant , linux-raid@vger.kernel.org List-Id: linux-raid.ids On Mon, 17 May 2010 17:38:04 -0700 Michael Evans wrote: > You should make it a sysfs controllable option to keep the counter > updated on spare discs or not. The default can be to keep it within > one as you currently are patching it to. Then later you can change > the default to the higher performance setting. Better, the presence > of that file can be used to test if the kernel supports it, and if it > doesn't initramfs scripts can react accordingly. There really isn't any need for user-space to know - this is an entirely in-kernel thing. mdadm already supports assembling arrays with widely different event_counts. It just gives them to the kernel and lets it make the final decision. And I think it is stretching things to call this a performance setting. It only changes behaviour when add devices is added/removed/failed or when an array is started or reshaped. None of those happen often enough that an extra spin-up is really going to be an issue. It is more an issue of simplicity and elegance. I don't think a sysfs option is appropriate for that. But thanks for the suggestion. NeilBrown