From owner-freebsd-fs@FreeBSD.ORG Fri Mar 4 13:18:52 2011 Return-Path: Delivered-To: freebsd-fs@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id E39FB106566B for ; Fri, 4 Mar 2011 13:18:52 +0000 (UTC) (envelope-from jdc@koitsu.dyndns.org) Received: from qmta06.emeryville.ca.mail.comcast.net (qmta06.emeryville.ca.mail.comcast.net [76.96.30.56]) by mx1.freebsd.org (Postfix) with ESMTP id C44988FC0A for ; Fri, 4 Mar 2011 13:18:52 +0000 (UTC) Received: from omta04.emeryville.ca.mail.comcast.net ([76.96.30.35]) by qmta06.emeryville.ca.mail.comcast.net with comcast id F1Jk1g0050lTkoCA61Js5K; Fri, 04 Mar 2011 13:18:52 +0000 Received: from koitsu.dyndns.org ([98.248.33.18]) by omta04.emeryville.ca.mail.comcast.net with comcast id F1Jp1g00c0PUQVN8Q1JpVg; Fri, 04 Mar 2011 13:18:50 +0000 Received: by icarus.home.lan (Postfix, from userid 1000) id 236189B422; Fri, 4 Mar 2011 05:18:49 -0800 (PST) Date: Fri, 4 Mar 2011 05:18:49 -0800 From: Jeremy Chadwick To: Tom Evans Message-ID: <20110304131849.GA26774@icarus.home.lan> References: <0E00DAFC-C39F-47DC-B9AF-16419C20997F@tao.org.uk> <615F1346-E830-42E2-B229-4181B8BC56BD@exonetric.com> <53FA69D2-2EF0-4CBF-985B-6E710F15FE02@tao.org.uk> <20110302001650.GB49147@icarus.home.lan> <27423168-85BE-41B1-8E14-94F01310EFE4@tao.org.uk> <20110302121612.GA61020@icarus.home.lan> <9EAE56CB-0CE9-4A08-B783-3EF9B1059E62@tao.org.uk> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: User-Agent: Mutt/1.5.21 (2010-09-15) Cc: freebsd-fs@freebsd.org, Dr Josef Karthauser Subject: Re: ZFS on a single disk? X-BeenThere: freebsd-fs@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Filesystems List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 04 Mar 2011 13:18:53 -0000 On Fri, Mar 04, 2011 at 12:13:58PM +0000, Tom Evans wrote: > On Fri, Mar 4, 2011 at 12:01 PM, Dr Josef Karthauser wrote: > > On 2 Mar 2011, at 12:16, Jeremy Chadwick wrote: > > > >> On Wed, Mar 02, 2011 at 10:28:50AM +0000, Dr Josef Karthauser wrote: > >>> > >>> I guess I should have been a bit clearer, I mean without any external > >>> devices backing them. Seeing as this is to be running on a VPS/Xen at > >>> a hosting environment I'm not able to add any additional devices. I'm > >>> not memory starved however. Is it madness to use a memory disk? (The > >>> instance in question will have 6gb of ram total to play with.). > >> > >> Again: don't disable the ZIL, and adjust the ARC to whatever size you > >> want using a combo of vm.kmem_size and vfs.zfs.arc_max.  Make use of the > >> RAM you have available to you.  :-) > >> > > > > Ok. That sounds like a plan. > > > > Next question, if I may. It looks like a XENHVM environment is unavailable > > to me, so I'm going have to stick with i386, and PAE to make the additional > > memory available. Given this constraint, does it still make sense to be > > trying to use ZFS? I had imagined that with 6GB to play with, I could > > dedicate 2GB of kernel space to the ZFS, leaving most of the rest for user > > land to play with. I've never played with PAE so I don't know whether I'm > > making any stupid assumptions here. > > > > Could I be, or is this still a sensible configuration to be thinking about? > > > > The grand picture is something like this then: > > > > - Hardware environment: XEN with 6GB ram, ample processor, 80GB disk. > > - O/S environment: FreeBSD 8.2 + PAE + XEN_PV + ZFS. > > - Goal: Host a number of ZFS managed jails in a secure and reliable way. > > > > Am I asking for trouble, or am I on the money? > > > > Joe > > > > p.s. there are some HVM XEN hosts out there, but it looks like they're all > > _much_ more expensive that the hosting provider that I'm considering, which > > appears to be offering a sweet spot of memory/cost. > > PAE? What is this, the 90s? ;) > > i386 + ZFS is not a wise move IMO. Any reason why you cannot use amd64? Not to mention, as I understand it, PAE induces all *sorts* of driver incompatibilities and technical hurdles when it comes to the kernel. I believe there are features/drivers/etc. which also do not build/work if PAE is used in the kernel. If something suddenly starts acting "oddly" on a PAE system, it wouldn't surprise me if PAE was to blame. Joe should be aware that amd64 does offer i386 compatibility libraries (referred to as "lib32") so you definitely can run i386 binaries on amd64. Speculative: Joe may be in the same boat I am with one of my customers: they use PHP eAccelerator (binary-only, only works on i386 *and* uses compat6x) as well as ZendOptimizer (works on amd64 but I'm not sure if this works with newer PHP), plus cannot upgrade to PHP 5.3 for a multitude of reasons. I also have no idea how the above two PHP extensions would play with a PHP built natively for amd64 (e.g. PHP amd64, eAcclerator i386 + compat6x libraries/libc/etc., ZendOptimizer amd64). I have fought time and time again with this customer about the use of this software ("is there REALLY any way you can get rid of it?"), but they absolutely will not budge. Anyway, you can see my predicament, but I'm not sure what Joe's is. I tend to build all of my systems with WITHOUT_LIB32=true in src.conf and without COMPAT_FREEBSD32 in the kernel, simply because I'd rather as much as possible of the system as possible be true 64-bit. Mix-matching the two has never sat well with me; probably the OCD part... :-) -- | Jeremy Chadwick jdc@parodius.com | | Parodius Networking http://www.parodius.com/ | | UNIX Systems Administrator Mountain View, CA, USA | | Making life hard for others since 1977. PGP 4BD6C0CB |