From owner-freebsd-performance@FreeBSD.ORG Thu Apr 8 15:43:12 2010 Return-Path: Delivered-To: freebsd-performance@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id B7FA1106564A for ; Thu, 8 Apr 2010 15:43:12 +0000 (UTC) (envelope-from ales@ripn.net) Received: from iruno.ripn.net (iruno.ripn.net [194.226.71.76]) by mx1.freebsd.org (Postfix) with ESMTP id 73A2F8FC16 for ; Thu, 8 Apr 2010 15:43:12 +0000 (UTC) Received: from [194.226.71.77] (helo=runo.ripn.net) by iruno.ripn.net with esmtp (RIPN) id 1Nztj0-000HRB-Ac for freebsd-performance@freebsd.org; Thu, 08 Apr 2010 19:32:58 +0400 Received: from ales by runo.ripn.net with esmtp (RIPN) id 1Nztj0-000HR8-5T for freebsd-performance@freebsd.org; Thu, 08 Apr 2010 19:32:58 +0400 From: "Alexandr D. Sergeev" To: freebsd-performance@freebsd.org Message-ID: <20100408153257.GA35557@runo.ripn.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline X-Virus-Scanned: ClamAV Date: Thu, 08 Apr 2010 19:32:58 +0400 Subject: Extreme performance falling on ZFS by durable operation of a server X-BeenThere: freebsd-performance@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Performance/tuning List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 08 Apr 2010 15:43:12 -0000 Hi, all! There are problems with a server on FreeBSD 7.2 CURRENT (6 Jul 2009), carrying out the task of mail server SMTP, IMAP. Messages are stored in maildirs on file system ZFS (mirror of 2 SAS 3k drives). Dovecot is used as IMAP server. After reboot the system works perfectly, however, somewhere in a month, at the same load, arise periodic long delays at opening of new letters. Judging by output iostat -x -w 1, during these moments the disk subsystem is strongly loaded. %busy disks comes nearer to 100 %. Readings nearby 300 per sec prevail. and approx. 10MBytes per sec. Somebody can advise, what diagnostics needed, for understanding of the reason of problems? There is an opinion that a problem with ZFS. Probably, it does not have enough memory. Because during the periods of occurrence of a problem, ARC is have less 300MB: vmstat -m ... solaris 1245732 274745K - 25872717949 16,32,64,128,256,512,1024,2048,4096 ... Or it is bug of ZFS implementation. -- Best regards, Alexandr D.Sergeev