Subject: Re: ext2fs is in the tree
To: Peter Svensson <firstname.lastname@example.org>
From: Warner Losh <email@example.com>
Date: 06/11/1997 12:12:30
In message <Pine.LNX.3.96.970611191831.12992Dfirstname.lastname@example.org> Peter Svensson writes:
: have no spare ones which could take over while cleaning up. The question
: is: how bad will ffs be after a crach while async-mounted?
While not directly NetBSD, my experience with FreeBSD leads me to
believe that you have about a .01% chance of major corruption on a
fairly quiet system, and about a 1-5% chance of major corruption on a
"loaded" system (eg make world). I'm not sure that this is directly
applicable to NetBSD, but expect similar results.
My tests were done on a FreeBSD system running current in April.
Major corruption was defined as needing to run newfs. I couldn't get
major corruption to happen when running 20 tests of light load, and
got it once when running things on heavy load (about 30 tries). I got
no major corruption when I was heavily loaded and running not async
(eg metadata sync). My test was to start whatever load I was placing
on it and then hit the reset button... I had full backups of the
system just in case, and I'd recommend others that try this do similar
things. Oh, this was on a PPro-200 with a AIC-7880 built in to a
narrow SCSI device. I didn't check to see if the contents of the
files were sane, merely if fsck would fail on the disk or not.