[Linux-cluster] crash during gfs_grow

[Bob Peterson]

On Mon, 2007-07-16 at 00:12 -0400, Zak Thompson wrote:
> Hello all, last night we were growing the gfs system and during the
> growth we had a panic and had to bounce the machines.  Now everything
> is back up and the lvm is showing the correct disk size  however when
> we run a df it is still showing the same disk space/usage/  well this
> isn’t good we now lost 1tb of new disk space.  Any suggestions on how
> to get this back?
> 
>  
> 
> Cheers,
> 
> Zak

Hi Zak,

What release of the software were you using when the panic occurred?
The reason I'm asking is because I'm trying to find out if you have
the relatively new fast df feature.  The fast df feature keeps track
of df changes differently than the older/slower df, which is more
accurate.  If you have this feature, you may want to temporarily
disable it and do a df to see if it shows better numbers.  It's fairly
new so at this point I'm going to assume you don't have that feature.

The problem of whether or not the newly added space is usable depends
on the state of the file system and whether the RGs were added
into the file system correctly, and that depends on how far it got
before the system panicked.

I recommend running gfs_fsck on the file system with a recent version
of gfs_fsck.  Recent versions of gfs_fsck are able to check the 
integrity of the RGs and possibly repair them if they're not in order.

If the RGs were not added in, you could probably just do another
gfs_grow to get them added in.

Regards,

Bob Peterson
Red Hat Cluster Suite