The boot runlevel attempts to fsck a file system that is not efs2. I have an all XFS system and the boot scripts run fsck instead of the xfs scripts - actually as I understand it nothing is needed for xfs as it does it on it's own.
Daniel, shouldnt fsck run the xfs utils if they are installed and its an XFS filesystem ?
The asumption that journalized filesystems (xfs in this case) do never need an fsck is not correct. In case the journals metadata is corrupt (Ive seen such cases even with xfs) it is required to run xfs_check. Now , the fsck utility is basically only a wrapper to the filesystems' native fsck (fsck.xfs in this case) and must be invoked on every boot. In case the filesystems metatdata information has been correctly committed upon mount (metadata checks do ocurr on fs mount), fsck.xfs will exit imediately and take no action. If the filesystem marks the metadata as being corrupt, fsck.xfs will perform a full filesystem check and eventually force single user mode requireing to run xfs_repair on the affected partition. IMO the behaviour of fsck being called on every boot is correct.
Oh, bleh, I read wrong .. thought it ran ext2 fsck, and not the xfs one ... Anyhow, Bjoern is correct, this happens even with ext3. Usually you would just set the amount of days to check to 0: # e2fstune -i 0 /dev/hd?? I do not know how to do this with xfs. Anyhow, marking invalid ...
