From owner-freebsd-stable Sat Nov 16 19:20:22 2002 Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 6827D37B401 for ; Sat, 16 Nov 2002 19:20:18 -0800 (PST) Received: from hub.org (hub.org [64.49.215.141]) by mx1.FreeBSD.org (Postfix) with ESMTP id EA2BD43E6E for ; Sat, 16 Nov 2002 19:20:17 -0800 (PST) (envelope-from scrappy@hub.org) Received: from hub.org (hub.org [64.49.215.141]) by hub.org (Postfix) with ESMTP id A9BA28A2D14; Sat, 16 Nov 2002 23:20:12 -0400 (AST) Date: Sat, 16 Nov 2002 23:20:12 -0400 (AST) From: "Marc G. Fournier" To: Peter Hoskin Cc: Lefteris Tsintjelis , Subject: Re: FreeBSD: Server or Desktop OS? In-Reply-To: <20021117122626.C301-100000@extortion.peterh.dropbear.id.au> Message-ID: <20021116221745.X23359-100000@hub.org> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG On Sun, 17 Nov 2002, Peter Hoskin wrote: > Because this is development of what will become stable. Always helps to > read the documentation. You know, this is great for those running a desktop machine, or small servers ... but really screws anyone that is running anything resembling a loaded server ... venus / jupiter are my pride and joy ... I've seen them hit loadavg's of 1000 and still be able to type (albeit very slowly) ... I've seen them hit 3500+ processes running ... I've had jupiter up to >150 jails running before I brought venus online ... and, until today, I've been running -STABLE on them ... I don't expect -STABLE to be rock solid all of the time ... I follow the commit messages, see what is changing ... I have a policy (or had until this weekend) that states for my customers: "As with any operating system, FreeBSD is a constantly evolving and changing entity, with changes made to improve performance, improve security, and improve stability being made continuously. As such, unless something critical has come up to override that, we try and upgrade our servers approximately every 30 days, in order to ensure that we do not diverge to greatly from what is considered the latest STABLE versions. " Just before 4.7-RELEASE was tag'd and put out, Matt Dillon spent many hours debugging a crash I was having on venus in the VM sub-system ... since RELEASE was so close, he didn't put it in until afterwards, so, of course, its only available in -STABLE ... this weekend, I downgraded venus/jupiter to RELENG_4_7, so now have lost that fix until 4.8 is released ... What happens when RELENG_4_7 crashes? I cvsup'd RELENG_4_7_0_RELEASE (or whatever its tag is) and then upgraded to RELENG_4_7, just to see what changes are in since 4.7 *was* released ... here is it: Updating collection src-all/cvs Edit src/Makefile.inc1 Edit src/UPDATING Edit src/contrib/bind/CHANGES Edit src/contrib/bind/bin/named/db_defs.h Edit src/contrib/bind/bin/named/db_sec.c Edit src/contrib/bind/bin/named/ns_defs.h Edit src/contrib/bind/bin/named/ns_ncache.c Edit src/contrib/bind/bin/named/ns_req.c Edit src/contrib/bind/bin/named/ns_resp.c Edit src/contrib/bind/lib/nameser/ns_name.c Edit src/contrib/bind/lib/nameser/ns_samedomain.c Edit src/contrib/tar/src/extract.c Edit src/contrib/tar/src/misc.c Edit src/crypto/heimdal/kadmin/version4.c Edit src/crypto/kerberosIV/kadmin/kadm_ser_wrap.c Edit src/kerberos5/include/version.h Edit src/kerberosIV/include/version.h Edit src/lib/libc/net/gethostbydns.c Edit src/lib/libc/net/getnetbydns.c Edit src/release/doc/de_DE.ISO8859-1/errata/article.sgml Edit src/release/doc/de_DE.ISO8859-1/hardware/alpha/article.sgml Edit src/release/doc/de_DE.ISO8859-1/hardware/alpha/proc-alpha.sgml Edit src/release/doc/de_DE.ISO8859-1/hardware/common/dev.sgml Edit src/release/doc/de_DE.ISO8859-1/hardware/i386/article.sgml Edit src/release/doc/de_DE.ISO8859-1/hardware/i386/proc-i386.sgml Edit src/release/doc/de_DE.ISO8859-1/installation/alpha/article.sgml Edit src/release/doc/de_DE.ISO8859-1/installation/i386/article.sgml Edit src/release/doc/de_DE.ISO8859-1/readme/article.sgml Edit src/release/doc/de_DE.ISO8859-1/relnotes/alpha/article.sgml Edit src/release/doc/de_DE.ISO8859-1/relnotes/common/new.sgml Edit src/release/doc/de_DE.ISO8859-1/relnotes/i386/article.sgml Edit src/release/doc/de_DE.ISO8859-1/share/sgml/catalog Edit src/release/doc/de_DE.ISO8859-1/share/sgml/release.dsl Delete src/release/doc/de_DE.ISO8859-1/share/sgml/release.ent Edit src/share/examples/cvsup/standard-supfile Edit src/sys/conf/newvers.sh Finished successfully Basically, nothing ... how many fixes are in -STABLE for bugs in the kernel that aren't "commonly triggered", but are like little time bombs that could be avoided? But, back to my question "What happens when RELENG_4_7 crashes?" ... in the past, before I did anything else, I'd upgrade to -STABLE, figuring it might be something that someone else caught and was fixed ... but, looking at the above changes in RELENG_4_7, it seems that reporting the crash is more or less useless ... if its already fixed in -STABLE, is someone going to MFC it down to RELENG_4_7? Up until this weekend, I had two heavily used/loaded servers pounding on -STABLE ... if one crashed, I had netdump in place to dump core to the other server, so that I had a crashdump ... and I'd report the results I could figure out, in hopes that *someone* would look at it and get it fixed, or ask me for more information on the bug ... basically, I'd risk a crash in the hopes of solidifying the OS for the next release, but I kind of hope(d) that -STABLE would at least run for more then a day or two :( Waiting for and upgrading to 4.8-PRERELEASE is useless, as there are times where it takes several weeks for it to crash ... > On Sun, 17 Nov 2002, Lefteris Tsintjelis wrote: > > > Date: Sun, 17 Nov 2002 03:19:28 +0200 > > From: Lefteris Tsintjelis > > To: Peter Hoskin > > Cc: Marc G. Fournier , Hununu , > > freebsd-stable@FreeBSD.ORG > > Subject: Re: FreeBSD: Server or Desktop OS? > > > > It sure is misleading. Why is it called -stable then? You would expect > > to stand up to its name. > > > > Regards, > > Lefteris Tsintjelis > > > > Peter Hoskin wrote: > > > > > > STABLE is still a development branch. The name is misleading. If I were > > > you, I'd install & run release. > > > > > > On Sat, 16 Nov 2002, Marc G. Fournier wrote: > > > > > > > On Sat, 16 Nov 2002, Hununu wrote: > > > > > > > > > On 16 Nov 2002 at 13:47, Geoffrey C. Speicher wrote: > > > > > > > > > > > > Am I expecting too much from FreeBSD-STABLE? Would I fair better if I > > > > > > > moved down into RELENG_4_7 and avoided -STABLE altogether? > > > > > > > > > > > > I think you're expecting too much from -stable. -stable is kind of a > > > > > > misnomer; read the Handbook section 21.2.2.1 ("What Is FreeBSD-STABLE?") > > > > > > for more. Your conclusion above is addressed there (spoiler: don't use > > > > > > -stable in production unless your test environment convinces you that it > > > > > > will work). > > > > > > > > > > Exactly. I have some boxes that run -stable. Though, I cvsup and install > > > > > world & kernel on one box I can afford to fail. If it works ok, then I > > > > > slowly move it to others.. -STABLE is not meant to be rock solid, and > > > > > running it on production servers can be time-consuming for various > > > > > reasons. > > > > > > > > The thing is, I run -STABLE on about a dozen desktops and servers at my > > > > day job, and have never had a problem ... but they don't carry near as > > > > much load as venus/jupiter do ... > > > > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-stable" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message