Corrupt filesystem

Vortex375 · 2010-07-01 19:00:36

Hi,

I recently made a fresh install on my laptop computer using ext4 as root filesystem. Today I ran an upgrade via 'pacman -Syu'. It installed KDE 4.4.5 among a few other things.

Afterwards, I wanted to suspend the machine to RAM - but for some reason it crashed.

Now the machine won't boot properly, because some files in /etc as well as a lot of .so files are empty. They all belong to packages that were upgraded prior to the crash (I think). The files exist but they contain no data.

Of course, fsck ran after automatically after reboot, told me it was "recovering journal" and that the "filesystem is clean". The filesystem in fact seems to be intact, but the files' contents are missing.

Luckily, pacman is still working, so I should be fine if I simply reinstall all installed packages. Is there a handy command that accomplishes this?

Can I configure ext4 to be more robust to crashes, to prevent such massive data loss in the future?

graysky · 2010-07-01 19:21:03

Look at your pacman log to see which packages to reinstall, or just reinstall all of them You can add the nodelalloc flag to your /etc/fstab for those ext4 partitions if you want.

Vortex375 · 2010-07-01 20:25:32

or just reinstall all of them

That's what I thought of doing, but unfortunately I don't know of a convenient way to list and reinstall all installed packages. Is there a script available that does this?

I googled for "nodelalloc" and found the mount option "alloc_on_commit". What is the difference between those two? I can't quite tell from what I've found so far...:(

loafer · 2010-07-01 20:48:04

There's a section in the Pacman Tips on a full reinstall, but I've never had to use it myself.

http://wiki.archlinux.org/index.php/Pac … _reinstall

Vortex375 · 2010-07-01 21:48:26

Thanks, I managed to recover the system.:)

However, during the recovery process I rebooted the system a few times and one time it crashed during startup and gave a similar error message like that time it crashed when I put it to sleep.

The kernel outputs something that looks like a call trace of various functions that are called acpi_something to the console, then locks up.

Maybe there's a hardware fault. Is there a way I can check my CPU for errors?

Unfortunately it seems the kernel's error message did not get logged anywhere as the system crashed right after it was displayed.

nv · 2010-07-02 16:04:14

Hello Vortex,

I have no solution to your problem, but I would like to point out that I have a similar problem since I reformatted my hd with the /home partition with ext4. I have never had problems with ext3, but since I reformatted the drive, my xfce4-panel configuration gets lost from time to time and every four or five days I need to run fsck manually on startup. So I am really interested in the solution to your problem.

Vortex375 · 2010-07-16 01:14:36

For now, I put "nodelalloc" under mount options in /etc/fstab. Not sure if it's a "solution" but at least I didn't notice any performance issues or other drawbacks so far.

Adding optimizations is nice but I think large-scale delayed allocations are very risky and data safety should be more important than performance.

wolfdogg · 2012-02-29 03:19:36

i know this is an older post, but i wanted to comment on

alloc_on_commit

i was doing some research after i just suffered data loss from UNC errors, possible bad sectors on the hard drive is the reason. i suffererd some data loss, and i suspect thats its primarily because of the way the ext4 filesystem was handing the errors. when the errors showed up, i rebooted, and ran 'fsck', once that came back fixed, i rebooted again. once i run a program that was accessing from corrupted areas in a terminal, the UNC hard drive errors would start popping up again. Its as if the journaling went bad somewhere. I havent had this problem with ext3, i feel like im back in the windows98 data loss days again, and its as if ext4 didnt like what was going on. Im thinking that im starting to see a trend of ext4 data loss problems.

Im searching for the most reliable file system i can find now. After your last comment NV, i think i will stick with ext3 because i was reading this article and the same thing was mentioned http://www.h-online.com/open/Ext4-data- … ews/112892, or was that you? :-)

anyhow, im installing, and ready to hit the 'write' button on cfdisk, so i made up my mind, im going back to ext3.

Last edited by wolfdogg (2012-02-29 04:21:09)

bernarcher · 2012-02-29 10:25:42

wolfdogg, you really do know about our necro thread policy. This thread is far too old for the audience you want to address. Please open a new one on your own and link back here.

Closing this one.

Arch Linux

#1 2010-07-01 19:00:36

Corrupt filesystem

#2 2010-07-01 19:21:03

Re: Corrupt filesystem

#3 2010-07-01 20:25:32

Re: Corrupt filesystem

#4 2010-07-01 20:48:04

Re: Corrupt filesystem

#5 2010-07-01 21:48:26

Re: Corrupt filesystem

#6 2010-07-02 16:04:14

Re: Corrupt filesystem

#7 2010-07-16 01:14:36

Re: Corrupt filesystem

#8 2012-02-29 03:19:36

Re: Corrupt filesystem

#9 2012-02-29 10:25:42

Re: Corrupt filesystem

Board footer