From owner-freebsd-fs@FreeBSD.ORG Fri Aug 22 12:40:43 2014 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [8.8.178.115]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id ACEF0BCA; Fri, 22 Aug 2014 12:40:43 +0000 (UTC) Received: from mailout05.t-online.de (mailout05.t-online.de [194.25.134.82]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (Client CN "mailout00.t-online.de", Issuer "TeleSec ServerPass DE-1" (verified OK)) by mx1.freebsd.org (Postfix) with ESMTPS id 6CD0B3B3C; Fri, 22 Aug 2014 12:40:43 +0000 (UTC) Received: from fwd24.aul.t-online.de (fwd24.aul.t-online.de [172.20.26.129]) by mailout05.t-online.de (Postfix) with SMTP id 9700B60C67B; Fri, 22 Aug 2014 14:40:35 +0200 (CEST) Received: from [192.168.119.33] (G-boWoZdghnLQwTr0jzKCGl-GozAeGj6JchsLSkxjCvypUtI0ziGONMols82j5pQ47@[84.154.101.219]) by fwd24.t-online.de with (TLSv1.2:ECDHE-RSA-AES256-SHA encrypted) esmtp id 1XKo8x-0sfyvg0; Fri, 22 Aug 2014 14:40:35 +0200 Message-ID: <53F73A39.9090000@freebsd.org> Date: Fri, 22 Aug 2014 14:40:25 +0200 From: Stefan Esser User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:31.0) Gecko/20100101 Thunderbird/31.0 MIME-Version: 1.0 To: "freebsd-fs@freebsd.org" Subject: Re: [Bug 187594] [zfs] [patch] ZFS ARC behavior problem and fix References: In-Reply-To: Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit X-ID: G-boWoZdghnLQwTr0jzKCGl-GozAeGj6JchsLSkxjCvypUtI0ziGONMols82j5pQ47 X-TOI-MSGID: 562569bf-0c89-4a1f-aabb-8e25e99aa673 Cc: Michael Jung X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.18-1 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 22 Aug 2014 12:40:43 -0000 Am 22.08.2014 um 13:53 schrieb Michael Jung: > I have been using the patches on a number of machines both physical and virtual with between > 8-64GB without issues. These include 10/10-STABLE/11-Current serving NFS/SMB and a poudriere > build box that is boot on UFS and ZFS for poudriere. > > Uptimes of 90+ days without issue. It has been a godsend on all machines even my desktop with > 8GB full X11 and numerous application + jails. > > No ZFS tuning. I hope that whatever further review moves forward so this can be committed. > > --mikej I second this request. I've also been using this patch on a system that required reboots to recover from low performance states that were not well handled without the patch. My specific use case is a development system (i.e. builds ports and the world nearly every day, often multiple times a day), which is also used as a home server for uncritical services (e.g. as DLNA server). Since I convert and cut DVB-S broadcasts for archival purposes on this system, it does process up to 50 GByte of video files (when I get around to it). The MPEG TS files are stored on a 1TB UFS2 disk, while waiting to be processed, while the system uses a 4*2TB raidz1 ZPOOL for all other purposes (including storage of the converted video files). (And yes, I know that a 4 disk raidz is not optimal, but the performance problems I observed without the patch are not caused by this ...) System specs: i7/2600 24GB RAM 4*2TB ZFS raidz1 1*1TB UFS2 Anyway, even the discouraged combination of ZFS and UFS works without any problem with Karl's patch. The system used to grind to a halt during files system operation (virtually freezing for some 20 to 30 seconds), which makes interactive use for video cutting and conversion rather painful (and does not help when streaming videos to a Smart-TV for display). I don't mind to apply the patch by hand, but since it offers massive performance improvements on systems that trigger the bad behaviour of the current code, this patch should be committed to -CURRENT and MFCed to at least 10-STABLE (and included in 10.1). People comparing different operating systems under load (and include FreeBSD with ZFS in this comparison) will else get disappointing results on out-of-the-box installations. We know from many reports (and in my case from personal experience), what big difference this patch makes! This patch should really have gone into -CURRENT, months ago, IMHO ... Regards, STefan