[Gluster-users] Troubles with sharding

[Stefan]

Hi, 

I have a lot of troubles with sharding. Firstly, all the shards are stored in the same directory, which can make GlusterFS quite slow if it's on a large HDD and the shard file size is small (e.g. 64MB). This can easily lead to >100k files, which is quite slow to run stats on, so GlusterFS will be very slow to operate. 

And then it's often the case that the size reporting on the FUSE mount is wrong, for example here: 
=============== 
# ls -la 
total 9223370277421359838 
drwx------ 1 root root 196 Jan 29 19:37 . 
drwx------ 1 root root 278 Jan 29 19:37 .. 
-rw-r----- 1 root root 1075 Jan 29 19:37 coverblacklist.frm 
-rw-r----- 1 root root 589824 Jan 29 14:20 coverblacklist.ibd 
-rw-r----- 1 root root 1564 Jan 29 19:37 cover.frm 
-rw-r----- 1 root root 879956852736 Jan 29 19:36 cover.ibd 
-rw-r----- 1 root root 1869 Jan 29 19:37 coverproperties.frm 
-rw-r----- 1 root root 536870912 Jan 29 14:21 coverproperties.ibd 
-rw-r----- 1 root root 61 Jan 29 19:37 db.opt 
=============== 
Note the total size (8 ExiBytes?), and the size for coverproperties.ibd is reported as 512MB, but it's actually 2.89GiB, I tested it with pv: 

# pv < coverproperties.ibd > /dev/null 
2,89GiB 0:00:36 [80,3MiB/s] [=====================================================================================================================================] 577% 

What can be the problem and the solution here? 

Thanks, 

Stefan