[Gluster-devel] unify: No space left on device

[Kevan Benson]

Albert Shih wrote:
>  Le 29/11/2007 à 19:42:55+0100, Albert Shih a écrit
>> Hi all
>>
>> I use glusterfs with unify volume and nufa scheduler.
>>
>> On each node (ie node10) I have :
>>
>> volume node10
>>   type storage/posix
>>   option directory /_glusterfs
>> end-volume
>>
>>
>> volume unify
>>   type cluster/unify
>>     subvolumes node01 ... node28
>>  option scheduler nufa
>>  option nufa.local-volume-name node10
>>  option nufa.limits.min-free-disk 5%
>>  option namespace ns
>> end-volume
>>
>> and it's working. When I've loggin node10 and write in my glusterfs mounted
>> partition it's write directly in my /_glusterfs partition. 
>>
>> But what's not very good is when the physical partition is full like 
>>
>> /dev/cciss/c0d1p1     133G  124G  2.1G  99% /_glusterfs
>>
>> and I put a file with size (for example) 5 Go glusterfs don't seem to able
>> to split the file in two. I've got the file system full error message.
> 
> I mean : one file until the file system is full and a second file on other
> node. 

Unify doesn't split files.  It just abstracts which server the file is 
stored on and retrieved from, and the aggregation of the files from all 
the servers when a listing is done.  What you should be seeing in this 
case, is that the *whole file* is written to a node that does have space 
(if there is a node that has more than 5% space available).

Is you used the striping translator above the unify, you would be 
storing chunks of files (if the files are large enough), and the chunks 
will be scattered across the nodes.  That would allow for more efficient 
space usage.

-- 

-Kevan Benson
-A-1 Networks