[Ocfs2-users] df & du - that old chestnut

[Sunil Mushran]

http://oss.oracle.com/~smushran/.debug/scripts/stat_sysdir.sh

Please file a bugzilla and attach the output of the above script.
http://oss.oracle.com/bugzilla

I think I know the issue. No, it is not related to blocksize/clustersize.

Sunil

Nigel Bishop wrote:
>
> Afternoon,
>
> We have an ocfs2 release 1.4 filesystem shared between two nodes 
> (RHEL5). The filesystem in question is used exclusively for Oracle 
> RMAN backups.
>
> A df –h shows the following:
>
> [root at imsthdb07 ~]# df -h /data/orabackup
>
> Filesystem Size Used Avail Use% Mounted on
>
> /dev/mapper/eva_mpio_myserver07_08_oracle_bkup0 250G 198G 53G 79% 
> /data/orabackup
>
> While a du –sh shows that 62G is used
>
> I read Sunil Mushran’s article on the reported differences between df 
> & du and that sort of makes sense (maybe I’m missing something). The 
> filesystem was created with a Blocksize and Clustersize of 4Kb and is 
> mounted with the following parameters:
>
> /dev/mapper/eva_mpio_myserver07_08_oracle_bkup0 /data/orabackup ocfs2 
> _netdev,nointr,defaults 0 0
>
> We have a RMAN retention policy of 3 days so backups older than that 
> do get deleted.
>
> As a test I did a df & du of this filesystem and then deleted (via 
> RMAN) the oldest backup, this file was 17Gb in size, after which I 
> again did a du & df. These both reported that 17Gb had been released 
> to the filesystem.
>
> What I still don’t get is why df reports that the Used space is at 
> 198Gb – that seems an awful waste
>
> Any insight into this major discrepancy would be much appreciated.
>
> Thanks,
>
> Nigel
>