Guaranteed filesystem corruption on unclean shutdown?
 

In the recent year, I was been unable to cleanly shutdown the system 5 times. In 4 of the times, the filesystem is severely corrupted such that most, if not all, files (inodes technical) become unusable. The cases happened on different Linux distribution & different kernel versions

Case 1: Debian Unstable x86_64, 2.6.18 kernel (custom build)
The power was unplugged accidentally after the computer was idle for several hours.
Result: after running e2fsck and blindly answering 'y' on all questions, ALL files and directories are moved under /lost+found...
ls /lost+found was beautiful (with highlight based on file type) :)

Case 2: RHEL 4 x86, vmware, 2.6.9 (vendor build)
This was a high loading VM. Several minutes after telling it to do some memory intensive work (needs much swapping), I found that the VM was rebooting. I don't know why, but I guess it was caused by hangcheck timer.
Result: Ran e2fsck. After answering 'y' several times, I gave up. And then many files could not found (e.g., /sbin/shutdown, so that I need another unclean shutdown)

Case 3: Fedora Core 6 x86_64, 2.6.18 kernel (vendor build)
I was running 'badblocks' on a logical partition, and using 'pup' to update software. The system loading should be low. Suddenly the computer became unresponsive. All keys (including Alt-F?, CapLock, Ctrl+Alt+Del) have no effect. The harddisk light flashed for several hours (which must be much longer than the badblocks needed). Eventually, I pressed Reset
Result: ls some directories gives ext2 errors (forget to copy). ls some files gives IO errors. When running e2fsck, I noted that it says something like 'Root directory is not a directory. Fix?'

Case 4: Fresh reinstall of Fedora Core 6 x86_64, 2.6.18 kernel (vendor build)
Computer hangs after completing installation process. Presed Reset
Result: same as about

It seems ext3 (or e2fsck) is extremely intolerant about power failure. I know one is supposed to cleanly umount before reboot. But it is always possible that power fails or machine hangs. I am really tired of waiting the thing to install and collecting traces of important data all over my harddisks and CD-ROMs (Don't tell me to backup. The point of this post is filesystem corruption, not data loss.) I am now typing this on Windows

Have anyone had similar experience? What can I do if the computer hangs or someone know nothing about computer unplug the power cable?

Hi

I am not sure if my experience is helpful.

I had similar situation when I had hardware failure. Disk drive was lousy. Chenged to new drive problem went away.

When case 2 or 3 happens, check syslog and dmesg.

Happy penguins!

I think ext3 is extremely tolerable.
My box hung many times because of my own stupidity and could only be recycled by power button on many occasions and I switched off UPS by mistake so power "failed" many times and never the filesystem was damaged...

I had a running domino server, samba, apache and other disk intensive programs.

But because I am not very experienced with Linux it is only IMHO.
Your problem could have to do with your hardwre setup. If it is the same machine.

I thought about hardware problem. But Case 2 happens in a VM. It seems unlikely to be hardware related.

Cases 1, 3, 4 happen on the same machine. 3 and 4 on the same day...

I live in a rural area where power interruptions are frequent (weather related). I use ext3, and I have never run e2fsck. On re-boot after power is restored, the ext3 journal is restored automatically.

Given the problems you describe, it's time to consider kernel-related problems, as in 'configuration'.

In /etc/sysconfig/kernel, there is this stanza:
Check your file to see if INITRD_MODULES includes ext3. If it is there, then should check the kernel config file in /usr/src/linux to see what is/is not compiled relative filesystem journaling.

The first question for the OP that I'd have is: What filesystem are you using? I'll assume ext2, and if so, I'd strongly suggest using a journaled filesystem such as ext3 or reiserfs.

Over the past couple of years, California has struggled with providing stable electricity, and unfortunately blackouts occur probably once or twice a year. Fortunately my machines have recovered (so far) without any problems or data loss under reiserfs. As a side note, making backups is always a good idea.

BTW, if you are using ext2 and want to convert it ext3, it's pretty easy to do

For what my experience is worth, ext3 shrugs aside such trivialities as power failures. On boot Debian just restores the filesystems and you'll miss this if you blink.

When I used ext2 I did have occasional glitches after power failures but, to be honest, nothing as bad as the OP describes.

How stupid I didn't mentioned ext3...

Today I tried again... repeated everything in Case 4.

But e2fsck says filesystem is clean after pressing reset :scratch: ... Using FC6 now :)