Thursday, October 05, 2006

Filesystem crash

On one of my boxes ext3 filesystem went down during updating system by yum. This appeared in log:

EXT3-fs error (device hda2): ext3_free_blocks_sb: \
bit already cleared for block 747938
Aborting journal on device hda2.
ext3_abort called.

with further info including – it was remounted to read-only mode and was slightly inconsistent. I reported this (a little bit freaky) to lkml and rebooted to correct the filesystem.

I tried to boot to single user mode, but fedora was (properly) faster and tried to check the root before I was able to log in. In spite of that it cannot correct it, so it gave me a shell to do-it-on-your-own. Ok, I ran fsck on root partition, it found some bad links, orphaned entries and block mismatches. Fsck was able to correct all those errors, then I checked S.M.A.R.T. output (with nothing suspicious) and rebooted.

Second phase of recovery began now. Rpm database and updated files needed to be corrected. Only 10 packages (including glibc libraries) were updated and the last one (the most affected) was gnome-vfs2. I removed all updated packages (there were both updated version and old versions) and verified if installed packages were allright, some was, some wasn't. Doesn't matter, I updated it again by yum, this time it went through successfully. Next nice experience :) – it works for now (not first, not last crash, since I use a -mm tree).

P.S. Jiri Kosina wrote on lkml list that he expereinced similar problems, so we will see if this is bug or not.


JJ said...

юп, линукс сукс! :D

Anonymous said...

He, on this shit never happend on Windows. :P

gep said...

I have absolutely no idea, what do you want to express :D. But I undertake you, if windows' file system crashes, no fsck will help you, not even god (=Bill Gates), what will you need then? You'll need to reinstall your system ;).