Re: [Jfs-discussion] JFS mount after unclean shutdown

[Dave K. <sh...@au...>]

On Sun, 2006-03-12 at 16:26 -0500, Ross Vandegrift wrote:
> Hello everyone,
> 
> Our Amanda backup servers use JFS on ~2TB RAID arrays as virtual tape
> archive space.  Under normal circumstances, the setup works great.
> 
> However, twice in the last six months or so, the filesystems have been
> uncleanly unmounted.  Once due to a kernel crash when the FC storage
> array was accidently disconnected, and once this weekend for an
> as-of-yet unknown reason.
> 
> When this happens, the filesystem is left in an unmountable state.
> Mount responds that jfs couldn't find a valid superblock.  A simple
> fsck fixes the issue in less than 60 seconds.  There's never been any
> data loss as a result of this.
> 
> But it's a bad situation - our backups failed the last two nights
> after the filesystem was unmountable.
> 
> We're using RH Enterprise Linux 3 on the JFS boxes with kernel 2.4.21-37.ELsmp
> and jfsutils-1.1.2-2.  Any ideas on what we can do to prevent needing
> this fsck?

By design, jfs requires fsck to be run against the volume to replay the
journal after being shutdown uncleanly.  This is normally done at boot
time by the init scripts if the entry in /etc/fstab is set up correctly.
The important parts of fstab entry are the type must be jfs and the last
field should be non-zero, usually 1 for the root file system, and 2 for
any other mount point.

-- 
David Kleikamp
IBM Linux Technology Center