TranscendLinux.com

Sometimes, ext3 journals go bad. Typically this will be evident in journal or dma errors in dmesg (I'll provide some examples next time I come across one of these). In order to resolve it, perform the following:
* Take the server down and boot it into the rescue environment (this means KVM time for support)... do NOT mount any of the partitions.
* fsck the affected partition.
* Remove the journal from the affected partition
* re-fsck the affected partition
* Readd the journal to the affected partition

fsck -yf /dev/hda3
tune2fs -O ^has_journal /dev/hda3
fsck -yf /dev/hda3
tune2fs -j /dev/hda3

At this point you are safe to mount the partitions, make sure everything looks ok, and if so then the server is ready to go!