From owner-freebsd-fs Tue Sep 23 11:10:29 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id LAA24751 for fs-outgoing; Tue, 23 Sep 1997 11:10:29 -0700 (PDT) Received: from roguetrader.com (brandon@cold.org [206.81.134.103]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id LAA24743 for ; Tue, 23 Sep 1997 11:10:21 -0700 (PDT) Received: from localhost (brandon@localhost) by roguetrader.com (8.8.5/8.8.5) with SMTP id MAA01082 for ; Tue, 23 Sep 1997 12:11:03 -0600 (MDT) Date: Tue, 23 Sep 1997 12:11:03 -0600 (MDT) From: Brandon Gillespie To: freebsd-fs@freebsd.org Subject: Known problems with async ufs? Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-fs@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Just curious, I have gotten the general impression that async UFS is unstable, simply because everybody says you really want to use sync UFS, which is horribly slower than async ufs. If your BOX is on a UPS, and you are generally assured that nothing is going to cause it to simply 'go down', why WOULDNT you want to run async on ALL filesystems? What problems are there with it? -Brandon Gillespie From owner-freebsd-fs Tue Sep 23 21:31:54 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id VAA28953 for fs-outgoing; Tue, 23 Sep 1997 21:31:54 -0700 (PDT) Received: from mexico.brainstorm.eu.org (root@mexico.brainstorm.fr [193.56.58.253]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id VAA28928 for ; Tue, 23 Sep 1997 21:31:49 -0700 (PDT) Received: from brasil.brainstorm.eu.org (brasil.brainstorm.fr [193.56.58.33]) by mexico.brainstorm.eu.org (8.8.4/8.8.4) with ESMTP id GAA03571 for ; Wed, 24 Sep 1997 06:31:47 +0200 Received: (from uucp@localhost) by brasil.brainstorm.eu.org (8.8.6/brasil-1.2) with UUCP id GAA18439 for freebsd-fs@FreeBSD.ORG; Wed, 24 Sep 1997 06:31:35 +0200 Received: (from roberto@localhost) by keltia.freenix.fr (8.8.7/keltia-2.10/nospam) id CAA20430; Wed, 24 Sep 1997 02:16:24 +0200 (CEST) Message-ID: <19970924021624.34229@keltia.freenix.fr> Date: Wed, 24 Sep 1997 02:16:24 +0200 From: Ollivier Robert To: freebsd-fs@FreeBSD.ORG Subject: Re: Known problems with async ufs? References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 0.84 In-Reply-To: ; from Brandon Gillespie on Tue, Sep 23, 1997 at 12:11:03PM -0600 X-Operating-System: FreeBSD 3.0-CURRENT Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk According to Brandon Gillespie: > UFS, which is horribly slower than async ufs. If your BOX is on a UPS, > and you are generally assured that nothing is going to cause it to simply > 'go down', why WOULDNT you want to run async on ALL filesystems? What > problems are there with it? Terry will no doubt give you many reasons why async UFS is bad (I even agree with most of them), why we should be using soft updates (which I really want to have but McKusick is still working on them I think) and/or LFS (I'd like to have it too) but my current experience for a lightly loaded machine on which I read mail and recompile CURRENT, many of my FS are async and never got a problem with them. /dev/sd0a on / (local) /dev/sd0s2e on /usr (local) /dev/sd0s2f on /usr/local (local) /dev/sd12e on /var (local) /dev/sd0s2g on /users (local) /dev/sd12f on /news (asynchronous, local, noatime, nosuid) /dev/sd0s2h on /src (asynchronous, local) (/usr/src) /dev/sd12h on /spare (asynchronous, local) (/home/ncvs) /dev/sd0s2d on /work (asynchronous, local) /dev/sd11a on /x (asynchronous, local, nosuid) /dev/sd12g on /y (asynchronous, local, noatime) (/usr/obj) procfs on /proc (local) /work/spare on /x/ftp/pub/FreeBSD/XFree86-3.3 (local, read-only) I get panics (not often even for CURRENT) and locks but still have to lose something. The risk is there, I know and accept it. Async /usr/obj helps "make world" a lot. -- Ollivier ROBERT -=- FreeBSD: There are no limits -=- roberto@keltia.freenix.fr FreeBSD keltia.freenix.fr 3.0-CURRENT #35: Sun Sep 21 19:28:07 CEST 1997 From owner-freebsd-fs Tue Sep 23 22:26:38 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id WAA01890 for fs-outgoing; Tue, 23 Sep 1997 22:26:38 -0700 (PDT) Received: from usr07.primenet.com (tlambert@usr07.primenet.com [206.165.6.207]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id WAA01884 for ; Tue, 23 Sep 1997 22:26:35 -0700 (PDT) Received: (from tlambert@localhost) by usr07.primenet.com (8.8.5/8.8.5) id WAA02832; Tue, 23 Sep 1997 22:26:33 -0700 (MST) From: Terry Lambert Message-Id: <199709240526.WAA02832@usr07.primenet.com> Subject: Re: Known problems with async ufs? To: brandon@roguetrader.com (Brandon Gillespie) Date: Wed, 24 Sep 1997 05:26:32 +0000 (GMT) Cc: freebsd-fs@FreeBSD.ORG In-Reply-To: from "Brandon Gillespie" at Sep 23, 97 12:11:03 pm X-Mailer: ELM [version 2.4 PL23] Content-Type: text Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > Just curious, I have gotten the general impression that async UFS is > unstable, simply because everybody says you really want to use sync > UFS, which is horribly slower than async ufs. If your BOX is on a UPS, > and you are generally assured that nothing is going to cause it to simply > 'go down', why WOULDNT you want to run async on ALL filesystems? What > problems are there with it? The reason behind this statement is based on recoverability in case of failure. If you can guard against failure (including automatic orderly shutdown by your IPS), then this shouldn't be a problem for you. The reality is that fsck can only deterministically recover from one failed metadata operation. Anything more than one, and you will probably end up with metadata errors. For example, if you had 10 metadata operations outstanding at a time, you would have only a 1 in 1024 chance of fsck "guessing right" and putting the FS into the state where it would have been had there not been a failure. If you have 10 operations outstanding during a crash, 1 to 10 files will be damaged. Probably 7, actually, based on 3 out of 7 of them being parent/child directory operations. Synchronus metadata writes guarantee the ordering of metadata updates do that at most one file will be damaged, and that damage will be recoverable (since there is only one way the operation can go: it's deterministic for one operation away from sync'ed state). Other ways of making the same ordering guarantees are delayed ordered writes (patent pending USL; EXT2FS uses a similar technology; I don't know if it's infringing), persistent cache writes (like the hardware NVRAM producs for SBUS machines, etc.) where the RAM that caches the metadata writes doesn't go away on failure, and Soft Updates (from the Ganger/Patt paper from the University of Kentucky). Of these, Synchronus writes are slowest, DOW is next slowest, Soft Updates is within 5% of memory speed for bursty traffic, and NVRAM is fastest (but requires special hardware and serious FS modifications). If you can guarantee an orderly shutdown on power failure because you have a daemon that supports your UPS and your UPS allows monitoring the line status (like an APC525, for example, but *not* an APC200, which doesn't have a serial port), then it's much safer to run async; of course, you are still screwed if your power supply fails, and you aren't in the Soft Updates or NVRAM cases. That answer the question? Regards, Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. From owner-freebsd-fs Wed Sep 24 00:07:31 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id AAA09509 for fs-outgoing; Wed, 24 Sep 1997 00:07:31 -0700 (PDT) Received: from palrel1.hp.com (palrel1.hp.com [156.153.255.235]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id AAA09494 for ; Wed, 24 Sep 1997 00:07:27 -0700 (PDT) Received: from postbox.india.hp.com (postbox.india.hp.com [15.10.45.1]) by palrel1.hp.com (8.8.6/8.8.5tis) with ESMTP id AAA12771; Wed, 24 Sep 1997 00:07:15 -0700 (PDT) Message-Id: <199709240707.AAA12771@palrel1.hp.com> Received: from localhost by postbox.india.hp.com with ESMTP (1.39.111.2/16.2) id AA038034586; Wed, 24 Sep 1997 12:33:06 +0530 To: Terry Lambert Cc: freebsd-fs@FreeBSD.ORG Subject: Re: Known problems with async ufs? In-Reply-To: Your message of "Wed, 24 Sep 1997 05:26:32 GMT." <199709240526.WAA02832@usr07.primenet.com> Date: Wed, 24 Sep 1997 12:33:06 +0530 From: A Joseph Koshy Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk >>>> "Terry Lambert" writes > Other ways of making the same ordering guarantees are delayed ordered I read (over a year back) that soft updates were being worked on by Keith Bostic(?). Anyone have any idea of further progress on this? > which doesn't have a serial port), then it's much safer to run async; > of course, you are still screwed if your power supply fails, and you > aren't in the Soft Updates or NVRAM cases. This doesn't offer protection from OS crashes though, so you are still exposed to some degree of risk. Koshy My Personal Opinions Only. From owner-freebsd-fs Wed Sep 24 01:30:59 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id BAA15039 for fs-outgoing; Wed, 24 Sep 1997 01:30:59 -0700 (PDT) Received: from schizo.dk.tfs.com (mail.trw.dk [195.8.133.123]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id BAA15032 for ; Wed, 24 Sep 1997 01:30:55 -0700 (PDT) Received: from critter.freebsd.dk (critter.dk.tfs.com [140.145.230.252]) by schizo.dk.tfs.com (8.8.7/8.7.3) with ESMTP id KAA27387; Wed, 24 Sep 1997 10:30:23 +0200 (MET DST) Received: from critter.freebsd.dk (localhost.dk.tfs.com [127.0.0.1]) by critter.freebsd.dk (8.8.7/8.8.7) with ESMTP id KAA00804; Wed, 24 Sep 1997 10:29:43 +0200 (CEST) To: A Joseph Koshy cc: Terry Lambert , freebsd-fs@freebsd.org Subject: Re: Known problems with async ufs? In-reply-to: Your message of "Wed, 24 Sep 1997 12:33:06 +0530." <199709240707.AAA12771@palrel1.hp.com> Date: Wed, 24 Sep 1997 10:29:42 +0200 Message-ID: <802.875089782@critter.freebsd.dk> From: Poul-Henning Kamp Sender: owner-freebsd-fs@freebsd.org X-Loop: FreeBSD.org Precedence: bulk In message <199709240707.AAA12771@palrel1.hp.com>, A Joseph Koshy writes: >>>>> "Terry Lambert" writes > >> Other ways of making the same ordering guarantees are delayed ordered > >I read (over a year back) that soft updates were being worked on by >Keith Bostic(?). Anyone have any idea of further progress on this? yes, it's almost here. Kirk is travelling at this time, but we should see it before Xmas. It will probably be an "extra" initially. -- Poul-Henning Kamp FreeBSD coreteam member phk@FreeBSD.ORG "Real hackers run -current on their laptop." From owner-freebsd-fs Wed Sep 24 07:32:25 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id HAA01507 for fs-outgoing; Wed, 24 Sep 1997 07:32:25 -0700 (PDT) Received: from shadows.aeon.net (root@shadows.aeon.net [194.100.41.1]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id HAA01492 for ; Wed, 24 Sep 1997 07:32:09 -0700 (PDT) Received: (from bsdfs@localhost) by shadows.aeon.net (8.8.7/8.8.3) id QAA29832 for freebsd-fs@FreeBSD.ORG; Wed, 24 Sep 1997 16:56:15 +0300 (EET DST) From: mika ruohotie Message-Id: <199709241356.QAA29832@shadows.aeon.net> Subject: Re: Known problems with async ufs? In-Reply-To: <19970924021624.34229@keltia.freenix.fr> from Ollivier Robert at "Sep 24, 97 02:16:24 am" To: freebsd-fs@FreeBSD.ORG Date: Wed, 24 Sep 1997 16:56:14 +0300 (EET DST) X-Mailer: ELM [version 2.4ME+ PL31 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > /dev/sd0a on / (local) > /dev/sd0s2e on /usr (local) > /dev/sd0s2f on /usr/local (local) > /dev/sd12e on /var (local) > /dev/sd0s2g on /users (local) > /dev/sd12f on /news (asynchronous, local, noatime, nosuid) > /dev/sd0s2h on /src (asynchronous, local) (/usr/src) > /dev/sd12h on /spare (asynchronous, local) (/home/ncvs) > /dev/sd0s2d on /work (asynchronous, local) > /dev/sd11a on /x (asynchronous, local, nosuid) > /dev/sd12g on /y (asynchronous, local, noatime) (/usr/obj) > procfs on /proc (local) > /work/spare on /x/ftp/pub/FreeBSD/XFree86-3.3 (local, read-only) general question... (pardon my possible ignorance) assuming one runs a filesystem /usr/src (noatime), wouldnt one need to run /usr and / (noatime) too to get the best "performance" from that flag? only filesystem that i could think i'd need to run (atime) is /var/mail but anyone knowing any others i should run such way too, tell me, since that' show my machines are not set up... i've been using my filesystems on either (async) (noatime) (async,noatime) for some time now, recently even using dma, in the past my machine did hang/crash relatively often coz of bad ram/cache i had, yet i never lost anything... now it doesnt crash anymore, tho. ofcourse, important filesystems / and /usr are just (noatime), others (async,noatime) except /var/mail just (async). and yes, i separate filesystems a lot, /usr/src /usr/ports /usr/obj /usr/tmp /usr /var /var/log /var/mail / /tmp /news /home all go into their own partition... (disadvantages?) i know i'd feel really pissed off to lose one of those (async) filesystems, but yet, everything i have in those are one way, or another, replaceable, thus, the gambler in me wants to run them as fast as possible... and if someone has some tests or anything for me to run, i might be able to do that. (i run always fresh current, hardly even older than two weeks) and while we talk about filesystems, _good_ journalling fs would be nice... mickey From owner-freebsd-fs Wed Sep 24 09:31:17 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id JAA08607 for fs-outgoing; Wed, 24 Sep 1997 09:31:17 -0700 (PDT) Received: from roguetrader.com (brandon@cold.org [206.81.134.103]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id JAA08590 for ; Wed, 24 Sep 1997 09:31:08 -0700 (PDT) Received: from localhost (brandon@localhost) by roguetrader.com (8.8.5/8.8.5) with SMTP id KAA05688; Wed, 24 Sep 1997 10:31:54 -0600 (MDT) Date: Wed, 24 Sep 1997 10:31:54 -0600 (MDT) From: Brandon Gillespie To: Terry Lambert cc: freebsd-fs@FreeBSD.ORG Subject: Re: Known problems with async ufs? In-Reply-To: <199709240526.WAA02832@usr07.primenet.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Wed, 24 Sep 1997, Terry Lambert wrote: > That answer the question? Very much so, thanks! Basically, with all of the down-play towards using an async FS it seemed as if there may be bugs within the actual async code, and not just the crash/power-fail situation problems. I personally am running off a smart-ups, and have a daemon monitoring it so it will shutdown properly, if need be. -Brandon Gillespie From owner-freebsd-fs Wed Sep 24 11:59:28 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id LAA18198 for fs-outgoing; Wed, 24 Sep 1997 11:59:28 -0700 (PDT) Received: from iafnl.es.iaf.nl (uucp@iafnl.es.iaf.nl [195.108.17.20]) by hub.freebsd.org (8.8.7/8.8.7) with SMTP id LAA18185 for ; Wed, 24 Sep 1997 11:59:23 -0700 (PDT) Received: by iafnl.es.iaf.nl with UUCP id AA10615 (5.67b/IDA-1.5 for freebsd-fs@FreeBSD.ORG); Wed, 24 Sep 1997 20:57:57 +0200 Received: (from wilko@localhost) by yedi.iaf.nl (8.8.5/8.6.12) id TAA00972; Wed, 24 Sep 1997 19:34:56 +0200 (MET DST) From: Wilko Bulte Message-Id: <199709241734.TAA00972@yedi.iaf.nl> Subject: Re: Known problems with async ufs? To: brandon@roguetrader.com (Brandon Gillespie) Date: Wed, 24 Sep 1997 19:34:56 +0200 (MET DST) Cc: tlambert@primenet.com, freebsd-fs@FreeBSD.ORG In-Reply-To: from "Brandon Gillespie" at Sep 24, 97 10:31:54 am X-Organisation: Private FreeBSD site - Arnhem, The Netherlands X-Pgp-Info: PGP public key at 'finger wilko@freefall.freebsd.org' X-Mailer: ELM [version 2.4 PL24 ME8a] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk As Brandon Gillespie wrote... > On Wed, 24 Sep 1997, Terry Lambert wrote: > > That answer the question? > > Very much so, thanks! > > Basically, with all of the down-play towards using an async FS it seemed > as if there may be bugs within the actual async code, and not just the > crash/power-fail situation problems. > > I personally am running off a smart-ups, and have a daemon monitoring it > so it will shutdown properly, if need be. UPS don't buy you much if the machine crashes/panics on it's own. Lets say due to a memory parity error, or hard I/O error in swap or something. My point is: it's not only power fails that can hurt you. Wilko _ ____________________________________________________________________ | / o / / _ Bulte email: wilko@yedi.iaf.nl http://www.tcja.nl/~wilko |/|/ / / /( (_) Arnhem, The Netherlands - Do, or do not. There is no 'try' ----------------------------------------------------------------------Yoda From owner-freebsd-fs Wed Sep 24 12:55:45 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id MAA22316 for fs-outgoing; Wed, 24 Sep 1997 12:55:45 -0700 (PDT) Received: from roguetrader.com (brandon@cold.org [206.81.134.103]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id MAA22304 for ; Wed, 24 Sep 1997 12:55:40 -0700 (PDT) Received: from localhost (brandon@localhost) by roguetrader.com (8.8.5/8.8.5) with SMTP id NAA06513; Wed, 24 Sep 1997 13:56:11 -0600 (MDT) Date: Wed, 24 Sep 1997 13:56:11 -0600 (MDT) From: Brandon Gillespie To: Wilko Bulte cc: tlambert@primenet.com, freebsd-fs@FreeBSD.ORG Subject: Re: Known problems with async ufs? In-Reply-To: <199709241734.TAA00972@yedi.iaf.nl> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Wed, 24 Sep 1997, Wilko Bulte wrote: > As Brandon Gillespie wrote... > > On Wed, 24 Sep 1997, Terry Lambert wrote: > > > That answer the question? > > > > Very much so, thanks! > > > > Basically, with all of the down-play towards using an async FS it seemed > > as if there may be bugs within the actual async code, and not just the > > crash/power-fail situation problems. > > > > I personally am running off a smart-ups, and have a daemon monitoring it > > so it will shutdown properly, if need be. > > UPS don't buy you much if the machine crashes/panics on it's own. Lets > say due to a memory parity error, or hard I/O error in swap or something. > > My point is: it's not only power fails that can hurt you. hrm, true Ok... what (if any) plans are there to make async at least as ''stable'' as ext2fs? I'm under the impression that ext2fs does something so it can recover from "bad things" better.. -Brandon From owner-freebsd-fs Wed Sep 24 13:24:19 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id NAA24400 for fs-outgoing; Wed, 24 Sep 1997 13:24:19 -0700 (PDT) Received: from bob.scl.ameslab.gov (bob.scl.ameslab.gov [147.155.137.254]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id NAA24394 for ; Wed, 24 Sep 1997 13:24:15 -0700 (PDT) Received: from bob.scl.ameslab.gov (localhost [127.0.0.1]) by bob.scl.ameslab.gov (8.8.7/8.8.5) with ESMTP id PAA06819; Wed, 24 Sep 1997 15:24:19 -0500 (CDT) Message-Id: <199709242024.PAA06819@bob.scl.ameslab.gov> X-Mailer: exmh version 2.0zeta 7/24/97 To: Brandon Gillespie cc: Wilko Bulte , tlambert@primenet.com, freebsd-fs@FreeBSD.ORG Subject: Re: Known problems with async ufs? In-reply-to: Your message of "Wed, 24 Sep 1997 13:56:11 MDT." Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Wed, 24 Sep 1997 15:24:19 -0500 From: Chris Csanady Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk >On Wed, 24 Sep 1997, Wilko Bulte wrote: > >> As Brandon Gillespie wrote... >> > On Wed, 24 Sep 1997, Terry Lambert wrote: >> > > That answer the question? >> > >> > Very much so, thanks! >> > >> > Basically, with all of the down-play towards using an async FS it seemed >> > as if there may be bugs within the actual async code, and not just the >> > crash/power-fail situation problems. >> > >> > I personally am running off a smart-ups, and have a daemon monitoring it >> > so it will shutdown properly, if need be. >> >> UPS don't buy you much if the machine crashes/panics on it's own. Lets >> say due to a memory parity error, or hard I/O error in swap or something. >> >> My point is: it's not only power fails that can hurt you. > >hrm, true > >Ok... what (if any) plans are there to make async at least as ''stable'' >as ext2fs? I'm under the impression that ext2fs does something so it can >recover from "bad things" better.. It's more or less as stable, but they just brainwash their users into thinking that it is safe. In fact my friend recently had his entire fs trashed. Oh well.. Hopefully, this will not be an issue for long considering Kirk's soft updates work. Chris From owner-freebsd-fs Wed Sep 24 14:44:34 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id OAA29950 for fs-outgoing; Wed, 24 Sep 1997 14:44:34 -0700 (PDT) Received: from sumatra.americantv.com (sumatra.americantv.com [207.170.17.37]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id OAA29944 for ; Wed, 24 Sep 1997 14:44:29 -0700 (PDT) Received: from right.PCS (right.PCS [148.105.10.31]) by sumatra.americantv.com (8.8.5/8.8.5) with ESMTP id QAA14012; Wed, 24 Sep 1997 16:44:08 -0500 (CDT) Received: (from jlemon@localhost) by right.PCS (8.6.13/8.6.4) id QAA09907; Wed, 24 Sep 1997 16:43:36 -0500 Message-ID: <19970924164335.39206@right.PCS> Date: Wed, 24 Sep 1997 16:43:35 -0500 From: Jonathan Lemon To: Brandon Gillespie Cc: Wilko Bulte , tlambert@primenet.com, freebsd-fs@FreeBSD.ORG Subject: Re: Known problems with async ufs? References: <199709241734.TAA00972@yedi.iaf.nl> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 0.61.1 In-Reply-To: ; from Brandon Gillespie on Sep 09, 1997 at 01:56:11PM -0600 Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Sep 09, 1997 at 01:56:11PM -0600, Brandon Gillespie wrote: > Ok... what (if any) plans are there to make async at least as ''stable'' > as ext2fs? I'm under the impression that ext2fs does something so it can > recover from "bad things" better.. Ah, but async FFS _is_ just as stable as async ext2fs. Actually, that's not quite true; I believe that async FFS still takes some more precautions than async ext2fs, so it may be slower. And no, ext2fs doesn't do anything that I know of that allows it to recover "better" than FFS. As far as performance, there is work underway to make sync FFS just as fast as async, without sacrificing crash-recovery reliability. -- Jonathan From owner-freebsd-fs Wed Sep 24 16:00:10 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id QAA05159 for fs-outgoing; Wed, 24 Sep 1997 16:00:10 -0700 (PDT) Received: from usr03.primenet.com (tlambert@usr03.primenet.com [206.165.6.203]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id QAA05121 for ; Wed, 24 Sep 1997 16:00:02 -0700 (PDT) Received: (from tlambert@localhost) by usr03.primenet.com (8.8.5/8.8.5) id PAA22462; Wed, 24 Sep 1997 15:59:39 -0700 (MST) From: Terry Lambert Message-Id: <199709242259.PAA22462@usr03.primenet.com> Subject: Re: Known problems with async ufs? To: brandon@roguetrader.com (Brandon Gillespie) Date: Wed, 24 Sep 1997 22:59:38 +0000 (GMT) Cc: wilko@yedi.iaf.nl, tlambert@primenet.com, freebsd-fs@FreeBSD.ORG In-Reply-To: from "Brandon Gillespie" at Sep 24, 97 01:56:11 pm X-Mailer: ELM [version 2.4 PL23] Content-Type: text Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > hrm, true > > Ok... what (if any) plans are there to make async at least as ''stable'' > as ext2fs? I'm under the impression that ext2fs does something so it can > recover from "bad things" better.. I don't want to restart this argument with the Linux/EXT2FS camp. However, the answer is it's fsck is more willing to accept "any known good state" as opposed to "the known good state the Fs would have been had it shutdown normally instead of abnormally". IMO (and this *is* MO, and the EXT2FS proponents have other O's), this is unsatisfactory. They addressed this somewhat by guaranteeing that metadata writes would be ordered. It's unclear whether this infringes on USL's Delayed Ordered Writes patent. The end result of DOW is not the same as the end result of NVRAM or Soft Updates; what it means is that multiple file contents can be corrupted, even though the metadata is set (*not* really restored) to a known good state. For example, I had a slow IDE device that would frequently lose all of /usr (it was on the root partition) when I had a crash, because I was doing a lot of manipulation of the directory block that contained it's directory entry. Your mileage may vary; a particular MTBF (Mean Time Between Failures) does not guarantee that you will have a failure in that time. The upshot is that async *is* as ''stable;; as ext2fs when it is mounted async (the historical default). If you want to make ffs come back as transparently, then make the fsck run twice after a crash, and it will recover equivalently "well". It's my recommendation that unless you are running extraordinary measures, like a UPS that can gracefully shutdown your machine for you, then you probably want to mount your ext2fs sync as well. This will have the side effect of making it look as "graceless" as ffs when a recovery is needed. Regards, Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. From owner-freebsd-fs Wed Sep 24 16:09:19 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id QAA05812 for fs-outgoing; Wed, 24 Sep 1997 16:09:19 -0700 (PDT) Received: from usr03.primenet.com (tlambert@usr03.primenet.com [206.165.6.203]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id QAA05804 for ; Wed, 24 Sep 1997 16:09:12 -0700 (PDT) Received: (from tlambert@localhost) by usr03.primenet.com (8.8.5/8.8.5) id QAA23263; Wed, 24 Sep 1997 16:08:56 -0700 (MST) From: Terry Lambert Message-Id: <199709242308.QAA23263@usr03.primenet.com> Subject: Re: Known problems with async ufs? To: ccsanady@bob.scl.ameslab.gov (Chris Csanady) Date: Wed, 24 Sep 1997 23:08:55 +0000 (GMT) Cc: brandon@roguetrader.com, wilko@yedi.iaf.nl, tlambert@primenet.com, freebsd-fs@FreeBSD.ORG In-Reply-To: <199709242024.PAA06819@bob.scl.ameslab.gov> from "Chris Csanady" at Sep 24, 97 03:24:19 pm X-Mailer: ELM [version 2.4 PL23] Content-Type: text Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > Hopefully, this will not be an issue for long considering Kirk's soft > updates work. Hopefully it will be a general soloution, and not a specific soloution like the SVR4 UFS soloution in Appendix A of the Ganger/Patt paper. If it's not a general soloution, then you will need to maintain one set of ufs code for ffs to use, and another for lfs and the others that currently rely on the ufs code as well. It will also mean that there are no hooks to guarantee transactions are idempotent (multiple atomic transactions considered as an all-or-nothing unit) for something like a user accessible transaction tracking system. Finally, it will mean that the problem was not solved as a list of order-dependency vectors, and it will just as much of a problem to add support to other FS's as it was to add to ffs. 8-(. Better to not rush him and to let him get it right... Regards, Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. From owner-freebsd-fs Wed Sep 24 16:21:54 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id QAA06603 for fs-outgoing; Wed, 24 Sep 1997 16:21:54 -0700 (PDT) Received: from gatekeeper.tsc.tdk.com (root@gatekeeper.tsc.tdk.com [207.113.159.21]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id QAA06591 for ; Wed, 24 Sep 1997 16:21:47 -0700 (PDT) Received: from sunrise.gv.tsc.tdk.com (root@sunrise.gv.tsc.tdk.com [192.168.241.191]) by gatekeeper.tsc.tdk.com (8.8.4/8.8.4) with ESMTP id QAA28166; Wed, 24 Sep 1997 16:20:03 -0700 (PDT) Received: from salsa.gv.tsc.tdk.com (salsa.gv.tsc.tdk.com [192.168.241.194]) by sunrise.gv.tsc.tdk.com (8.8.5/8.8.5) with ESMTP id QAA18608; Wed, 24 Sep 1997 16:20:02 -0700 (PDT) Received: (from gdonl@localhost) by salsa.gv.tsc.tdk.com (8.8.5/8.8.5) id QAA12833; Wed, 24 Sep 1997 16:20:01 -0700 (PDT) From: Don Lewis Message-Id: <199709242320.QAA12833@salsa.gv.tsc.tdk.com> Date: Wed, 24 Sep 1997 16:20:01 -0700 In-Reply-To: Brandon Gillespie "Re: Known problems with async ufs?" (Sep 24, 10:31am) X-Mailer: Mail User's Shell (7.2.6 alpha(3) 7/19/95) To: Brandon Gillespie , Terry Lambert Subject: Re: Known problems with async ufs? Cc: freebsd-fs@FreeBSD.ORG Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk On Sep 24, 10:31am, Brandon Gillespie wrote: } Subject: Re: Known problems with async ufs? } I personally am running off a smart-ups, and have a daemon monitoring it } so it will shutdown properly, if need be. I've got four machines fed by an UPS. Within 48 hours of being switched to UPS power, someone working on an adjacent rack accidentally kicked loose the power cord on the output side of the UPS :-( You're also not protected against power supply failures and failures on the output side of the UPS (blow fuse, etc.). From owner-freebsd-fs Wed Sep 24 21:28:50 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id VAA23593 for fs-outgoing; Wed, 24 Sep 1997 21:28:50 -0700 (PDT) Received: from palrel1.hp.com (palrel1.hp.com [156.153.255.235]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id VAA23581 for ; Wed, 24 Sep 1997 21:28:35 -0700 (PDT) Received: from postbox.india.hp.com (postbox.india.hp.com [15.10.45.1]) by palrel1.hp.com (8.8.6/8.8.5tis) with ESMTP id VAA18568; Wed, 24 Sep 1997 21:28:13 -0700 (PDT) Message-Id: <199709250428.VAA18568@palrel1.hp.com> Received: from localhost by postbox.india.hp.com with ESMTP (1.39.111.2/16.2) id AA112501429; Thu, 25 Sep 1997 09:53:49 +0530 To: Terry Lambert Cc: brandon@roguetrader.com (Brandon Gillespie), wilko@yedi.iaf.nl, freebsd-fs@FreeBSD.ORG, koshy@postbox.india.hp.com Subject: Re: Known problems with async ufs? In-Reply-To: Your message of "Wed, 24 Sep 1997 22:59:38 GMT." <199709242259.PAA22462@usr03.primenet.com> Date: Thu, 25 Sep 1997 09:53:49 +0530 From: A Joseph Koshy Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk >>>> "Terry Lambert" writes > They addressed this somewhat by guaranteeing that metadata writes would > be ordered. It's unclear whether this infringes on USL's Delayed > Ordered Writes patent. EXT2FS has been around for a long time. Wouldn't the DOW patent be invalid on the grounds of ``prior art'' if the technology was similar? Just curious. Koshy My Personal Opinions Only. From owner-freebsd-fs Thu Sep 25 00:00:18 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id AAA01134 for fs-outgoing; Thu, 25 Sep 1997 00:00:18 -0700 (PDT) Received: from usr03.primenet.com (tlambert@usr03.primenet.com [206.165.6.203]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id AAA01127 for ; Thu, 25 Sep 1997 00:00:15 -0700 (PDT) Received: (from tlambert@localhost) by usr03.primenet.com (8.8.5/8.8.5) id XAA08542; Wed, 24 Sep 1997 23:53:43 -0700 (MST) From: Terry Lambert Message-Id: <199709250653.XAA08542@usr03.primenet.com> Subject: Re: Known problems with async ufs? To: koshy@india.hp.com (A Joseph Koshy) Date: Thu, 25 Sep 1997 06:53:43 +0000 (GMT) Cc: tlambert@primenet.com, brandon@roguetrader.com, wilko@yedi.iaf.nl, freebsd-fs@FreeBSD.ORG, koshy@postbox.india.hp.com In-Reply-To: <199709250428.VAA18568@palrel1.hp.com> from "A Joseph Koshy" at Sep 25, 97 09:53:49 am X-Mailer: ELM [version 2.4 PL23] Content-Type: text Sender: owner-freebsd-fs@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > > They addressed this somewhat by guaranteeing that metadata writes would > > be ordered. It's unclear whether this infringes on USL's Delayed > > Ordered Writes patent. > > EXT2FS has been around for a long time. Wouldn't the DOW patent be > invalid on the grounds of ``prior art'' if the technology was similar? No. 1993. > Just curious. No problem. DOw was not in the initial implementation. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.