A reiserfs catastrophe?
From: Chris Carlen (crobc_at_BOGUSFIELD.sbcglobal.net)
Date: 01/30/05
- Next message: Eric Haase: "Re: Wireless internet concerns"
- Previous message: Dances With Crows: "Re: Please explain fs_freq in /etc/fstab"
- Next in thread: mjt: "Re: A reiserfs catastrophe?"
- Reply: mjt: "Re: A reiserfs catastrophe?"
- Reply: David Efflandt: "Re: A reiserfs catastrophe?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Sat, 29 Jan 2005 18:09:27 -0800
Hi:
One of my Suse 9.1 boxes failed a couple days ago. I was using a single
40GB reiserfs partition for /, and a second reiserfs partition of 35GB
which contained all our digital photos. I had backed up the digital
photos, so all is not lost. Only confidence in either my computer or
OS, but I'm not sure yet which to blame.
The failure occurred at the moment I accessed the machine via NFS from
my other Suse 9.1 box, as I clicked on a JPEG file in my Konqueror
window, displaying the contents of a directory of the photos partition
on the defunct machine.
At that moment, the machine black-screened, and my machine's Konqueror
hung as it always does when an NFS server dies.
I was shocked, and immediately proceeded to see if only the X server of
that machine crashed. It is my wife's PC, and she was running KDE on it
while I accessed the NFS shares, though she wasn't actually doing
anything at that moment.
I pinged it, and sure enough it responded! Linux was still alive! But
I couldn't do a ssh to it.
Only after a few days of thinking about it did I realize that made
perfect sense, for a failed drive/filesystem, but not a Linux crash.
I couldn't reboot it or get any response but ping, so I turned it off.
Then I ran the Suse rescue CD, mounted the drive, and began seeing what
was left of it. It wouldn't mount, but fsck was able to put it back
together somewhat.
What was left was a long list of numbered directories and numbered files
in lost+found. Fortunately, much of my Photoshop work was still there
(I hadn't yet backed up this stuff, only the original RAW files from my
DSLR camera). Even the filenames were intact. Only most of the
organization was trashed. Also about 95% of my wife's junk was
recovered. Not much of it was important enough to her to make her ever
respond to my admonishments that she should regularly backup, even
though I made an automagic script for her to do it. But it didn't
matter anyway. She still has her junk.
Unfortunately, the system log files couldn't be found. I was hoping to
see if there was some last critical clue as to what went wrong, but not.
Western Digital's utility for checking their Caviar drives indicated
there was nothing wrong with the drive. I suppose we will simply
reinstall Linux on it and start over.
At this point there really seems to be very little that could lead us to
be certain if the fault was a reiserfs bug or just a hardware glitch.
The fact that the Linux kernel was still alive was interesting, but I'm
not sure it is conclusive evidence. I can imagine that a glitch in the
disk subsystem could result in disk corruption and making it no longer
accessible by the kernel, while not crashing the kernel which would
still be purring along in RAM.
Bummer there is no way to really know.
But I am hesitant to use reiserfs again. I tried it because my reserach
indicated it might perform a little bit better, for a "desktop" type of
system which entails mostly accessing many little files, rather than a
few large files which I heard ext2(3) is better for.
But for all practical purposes, I don't know if the performance
differences are really distinguishable to a user.
Also, a recent search on the Suse mailing list revealed that almost all
complaints about problems with filesystems were about reiserfs.
I think I will go with ext2 this time.
Comments welcome.
Good day!
-- _____________________ Christopher R. Carlen crobc@sbcglobal.net SuSE 9.1 Linux 2.6.5
- Next message: Eric Haase: "Re: Wireless internet concerns"
- Previous message: Dances With Crows: "Re: Please explain fs_freq in /etc/fstab"
- Next in thread: mjt: "Re: A reiserfs catastrophe?"
- Reply: mjt: "Re: A reiserfs catastrophe?"
- Reply: David Efflandt: "Re: A reiserfs catastrophe?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|