From owner-freebsd-questions Sun Feb 17 8:21:53 2002 Delivered-To: freebsd-questions@freebsd.org Received: from uni-sb.de (uni-sb.de [134.96.252.33]) by hub.freebsd.org (Postfix) with ESMTP id EF28537B402 for ; Sun, 17 Feb 2002 08:21:47 -0800 (PST) Received: from cs.uni-sb.de (cs.uni-sb.de [134.96.252.31]) by uni-sb.de (8.12.2/2002020400) with ESMTP id g1HGLjj06273; Sun, 17 Feb 2002 17:21:45 +0100 (CET) Received: from mail.cs.uni-sb.de (IDENT:9kl/W3fD33JTZfAa08iBFUHUnqWvWqxQ@mail.cs.uni-sb.de [134.96.254.200]) by cs.uni-sb.de (8.12.1/2001121800) with ESMTP id g1HGLis09211; Sun, 17 Feb 2002 17:21:44 +0100 (CET) Received: from wjpserver.cs.uni-sb.de (wjpserver.cs.uni-sb.de [134.96.247.42]) by mail.cs.uni-sb.de (8.12.2/2002011500) with ESMTP id g1HGLgO01609; Sun, 17 Feb 2002 17:21:43 +0100 (CET) Received: (from schuerge@localhost) by wjpserver.cs.uni-sb.de (8.11.0/8.11.0/wjp-SVR4/1999052600) id g1HGLgl27185; Sun, 17 Feb 2002 17:21:42 +0100 (MET) From: Thomas Schuerger Message-Id: <200202171621.g1HGLgl27185@wjpserver.cs.uni-sb.de> Subject: Re: New dir layout structure slow on full disks? In-Reply-To: <3C6FD5F7.7060702@potentialtech.com> from Bill Moran at "Feb 17, 2002 11:10:31 am" To: Bill Moran Date: Sun, 17 Feb 2002 17:21:42 +0100 (MET) Cc: freebsd-questions@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL57 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG > > Hi! > > > > I'm using 4.5-RELEASE and am quite happy with the new directory layout > > stuff. It really speeds up file scans etc. a lot. > > > > But there can be serious performance loss: > > > > When a disk is nearly full (e.g. 98%, but still 1,6 GB free, values > > taken from "df"), creating files becomes very slow. I noticed this > > when I wanted to install a big port (e.g. kdebase2). The extraction of > > the .tar.gz archive took really long and made my system more or less > > unusable. tar took something like 90-99% system time for 3 minutes. > > Err ... have you compared this to performance on a filesystem that's > close to full with the old dirpref code? I'm gussing it would be > similarly slow. > > ffs doesn't run well with close-to-full drives, it was never intended > to. I believe both the man pages for newfs and tunefs state that > performance drops exponentially on a filesystem as it passes the > 92% capacity mark. That's why, under normal conditions, only root > can fill it beyond that mark. > > When you say 98% full, do you mean 98% of the 92%, or do you mean > 98% of the total fs capacity? > > If the former is the case, then you really do have an issue that > needs looked at. If the latter is the case, then you are simply > reiterating a known limitation of ffs that's been there since (I > believe) BSD 4.2. I mean the user-visible percentage (that's why I said "values taken from df"). I didn't compare it to the performance of the old dirpref code, but I have never noticed anything similar before 4.5-RELEASE even if the disks where even more full (df-capacity >100%). When I deleted some files on the disk (from 98% to 96% capacity), creating files seemed fine again as ever. Greetings, Thomas. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message