[Cluster-devel] small inconsistency presenting fs stats between GFS1 and GFS2

[Fabio Massimo Di Nitto <fabbione@ubuntu.com>]

Wendy Cheng wrote:
> Fabio Massimo Di Nitto wrote:
>> Hi guys,
>>
>> this is purely cosmetic and I didn't prepare a patch but see this:
>>
>> given a 4GB block device (just as an example):
>>
>> /dev/nbd2             3,9G  518M  3,4G  14% /mnt/gfs2
>> /dev/nbd1             3,1G   20K  3,1G   1% /mnt/gfs
>>
>> you can see that gfs1 masks the device size to hide the journals, while gfs2
>> shows the journal as "in use" space.
>>
>> would it be possible to make gfs1 consistent with gfs2 (and probably the rest of
>> the world) by reporting the data in the same way? is it possible to do it
>> without breaking anything?
>>
>>   
> I would think it is ok but this is arguable.

I mean no flamewar :)

> Say in ext3 case, if the 
> journal is on an external device, will you count it as "in use" space ?

No, not really. it's not on the same device.

> I never heard our users complain about this (maybe until now ? :) ). 

It's more curiosity rather than a complain. My first concern looking at df -h
was (so to speak because I knew it was due to the journals): "3.1GB device? but
I allocated 4..."

> Changing this may cause some confusions with GFS1's existing 
> installation base (say customers may start to complain.. after RHEL x.y, 
> we start to see diskspace usage jump or something like that ?).

The free disk % won't change and yes, I understand that you would see more disk
in use, but also the real device size would increase and maintain the same balance.
Anyway I agree that it could be confusing with both approaches.

Thanks
Fabio

-- 
I'm going to make him an offer he can't refuse.