[Gluster-users] used space not reclaimed on file delete

[Joe Julian]

On May 4, 2014 9:04:58 PM PDT, Michael Cronenworth <mike at cchtml.com> wrote:
>On 05/04/2014 12:15 AM, Vijay Bellur wrote:
>> Do you notice this discrepancy when you perform deletions on a
>glusterfs client
>> mount?
>
>No.
>
>>
>> Deleting files directly from the glusterfs brick directories is not a
>> recommended practice.
>
>Is there a "heal" mechanism for non-replicated bricks that could
>scan/fix 
>directly deleted files? I have an app running on the server that can
>delete 
>data. Would the best practice in this case be to mount the brick on the
>server 
>and have the app talk that way?

That's not just best practice, that's documented design. Just like you wouldn't update a file by writing directly to the disk partition, you don't update a clustered volume by going directly to the brick. 

Always go through a client mount or the api library.