[Linux-cluster] gfs 5.1 question

[Bob Peterson]

On Thu, 2008-02-21 at 18:13 -0500, chris barry wrote:
> Ok, a question for the old-timers:
> 
> gfs 5.1
> GULM
> RH 7.3, custom kernel
> ~350GB FC array.
> Eg. ancient stuff - can't change it. migrating to new gfs cluster I just
> built soon.
> 
> df shows 84% of volume used, but writes say out of space.
> 
> What tests can I do to determine why this is, and better how can I
> eliminate this unwritable 16%? Do I need to 'reclaim' inodes or
> something?
> 
> 
> Thanks,
> -C

Hi Chris,

My guess is that the 16% is tied up in free metadata.
You may want to do a "gfs_tool reclaim" to turn the free metadata into
free data blocks.  (And as always, those will in turn be turned back
into metadata blocks as needed.)

Regards,

Bob Peterson
Red Hat GFS