Date: Wed, 14 Sep 2005 08:53:08 -0700 From: Brooks Davis <brooks@one-eyed-alien.net> To: Robert Watson <rwatson@FreeBSD.org> Cc: Alexey Dokuchaev <danfe@FreeBSD.org>, src-committers@FreeBSD.org, cvs-src@FreeBSD.org, cvs-all@FreeBSD.org, Marcel Moolenaar <marcel@xcllnt.net>, Garance A Drosehn <gad@FreeBSD.org> Subject: Re: cvs commit: src/sys/conf kern.post.mk Message-ID: <20050914155307.GA32734@odin.ac.hmc.edu> In-Reply-To: <20050914100957.L33820@fledge.watson.org> References: <20050911002229.51F4916A471@hub.freebsd.org> <432382BC.5080105@root.org> <200509110310.36423@harrymail> <43255152.3000609@freebsd.org> <20050912165518.GA94181@dragon.NUXI.org> <20050913125820.GA10663@FreeBSD.org> <F74CCD5E-AD4C-4343-9626-5F6460C4D6C6@xcllnt.net> <p06230916bf4d11bb5363@[128.113.24.47]> <20050914100957.L33820@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--n8g4imXOkfNTN/H1 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Wed, Sep 14, 2005 at 10:17:53AM +0100, Robert Watson wrote: > On Tue, 13 Sep 2005, Garance A Drosehn wrote: >=20 > >I must admit I'm a little uneasy moving /usr and /var into '/', just=20 > >because I'm so used to the way it is. I really *like* having /var as a= =20 > >separate partition. But hard disks are huge compared to how they used= =20 > >to be, so I don't mind having a 500-meg '/'. I *do* save the=20 > >kernel.debug information in my /boot/kernel's, and my root partition is= =20 > >still under 50% full. >=20 > We probably ought to move this discussion to another list, but -- I=20 > remember two very specific occasions where I first realized how important= =20 > an isolated /var is: >=20 > (1) In about 1995, when I first started using ppp, I discovered the > hard way that the default logging level was set a bit high, and filled > the entire hard disk with log records in a couple of days. Don't > remember which FreeBSD revision that was. It's amazing how little > time it takes, if you get a log messsage ever second or so, to fill up > a hard disk. I've also hit cases where the log come in so fast that you can fill a 1GB /var fast enough that newsyslog never has a chance to compress the log. My particular case appears to be a program trying to dump core when not fully paged in and when the executable was on nfs, but has been deleted. I've trained my users not to delete there executables so fast and that seem to have stopped the problems, but I haven't had a chance to find a way to trigger the bug in the lab. I've found it on 4.x, but wouldn't be suprised if it effects all systems. The killer is that it generates two log messages per fault as fast as it can fault so while the file compresses very nicely, syslog's repeated line compression doesn't work. :( -- Brooks --=20 Any statement of the form "X is the one, true Y" is FALSE. PGP fingerprint 655D 519C 26A7 82E7 2529 9BF0 5D8E 8BE9 F238 1AD4 --n8g4imXOkfNTN/H1 Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.1 (GNU/Linux) iD8DBQFDKEdjXY6L6fI4GtQRAlgIAJ9OmCIsiZmd9F4P39TFS+Dx9Mn/JACeNUMv TQwy4BJTiFaf644xM6im/Ew= =G5CV -----END PGP SIGNATURE----- --n8g4imXOkfNTN/H1--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050914155307.GA32734>