From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sven Riedel Subject: Re: SATA command timeouts Date: Mon, 10 Jul 2006 13:25:51 +0200 Message-ID: <20060710112551.GB14003@localnet> References: <20060708130037.GA22725@localnet> Reply-To: sr@gimp.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from mail.kamp-dsl.de ([195.62.99.42]:58552 "HELO mailout.kamp-dsl.de") by vger.kernel.org with SMTP id S1751320AbWGJLZy (ORCPT ); Mon, 10 Jul 2006 07:25:54 -0400 Received: from dsl-mail.kamp.net (mail.kamp-dsl.de [195.62.99.42]) by sponts.dsl-backup.kamp.de (SPONTS v6.3.6) with SMTP id 110C582FA64C-0 for ; Mon, 10 Jul 2006 11:25:53 +0000 Received: from mx01.kamp-dsl.de (mail.kamp-dsl.de [195.62.99.42]) by sponts.dsl-backup.kamp.de (SPONTS v6.3.6) with ESMTP id 110C582FA5DF-0 for ; Mon, 10 Jul 2006 11:25:53 +0000 Received: from mail.localnet (turing.localnet [192.168.13.1]) by mx (Postfix) with ESMTP id AE59B1F864 for ; Mon, 10 Jul 2006 11:25:54 +0000 (UTC) Content-Disposition: inline In-Reply-To: <20060708130037.GA22725@localnet> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org Cc: linux-ide@vger.kernel.org On Sat, Jul 08, 2006 at 03:00:37PM +0200, Sven Riedel wrote: > Hi, > I hooked up new SATA disks to a machine yesterday and set up a software raid. > After copying roughly 25 GB of data onto the array all attached drives > reported command timeouts and wouldn't recover. Only a reboot would help. > I tried the same again, this time the drives would timeout after 60 GB of data > written. I exchanged the CPU, but the same thing occurs. Well, seems the second CPU was defunct. After switching back to the first CPU the hard system crashes stopped occurring. > My > next test will be doing write-heavy operations on all four drives > simulaneously, but without gathering them into a raid array. Well, it looks like the problem can be reproduced most reliably only when all four drives are under load. It happens even sooner when they are gathered into a raid array (I only tested level 5 - could be that the background "rebuild" after creation is adding additional load which triggers the problem sooner). Just to rule out an insufficient power supply, I hooked the drives up to a completely independent PSU and the drives still all suddenly trigger a command-timeout after a while. Regs, Sven -- Sven Riedel sr@gimp.org