[Gluster-users] brick disk usage variance

Haychoi Taing taingh at u.washington.edu
Mon Nov 26 22:37:12 UTC 2012


I have a 5x3 distributed-replicate gluster v3.3.1 volume on xfs bricks 
running rhel6.3/x86_64. I've populated the volume with ~130G of data so 
far and noticed that triplicate bricks consume different amounts of disk 
space. Some triplicates have 2 of 3 bricks consume similar disk (1,2,3 
triplicate below), 4th triplicate all different sizes, while the 5th 
triplicate all consume the same amount of disk.

Host   Used
node1  19778532
node2  19778644   1
node3  19778532

node4  33153484
node5  33154232   2
node6  33154232

node7  18017708
node8  18017708   3
node9  18017692

node10 28565512
node11 28565416   4
node12 28565464

node13 34858100
node14 34858100   5
node15 34858100

Some more info... I started with just a replicated volume (1x3) then 
added each triplicate followed by 'gluster volume rebalance gv0 start' 
to end up with a 5x3 distributed-replicated volume. One more thing, I 
have 'cluster.quorum-type: auto'. 'gluster volume heal gv0 full' runs 
successfully and doesn't change disk usage.

Is brick disk consumption variance normal?


More information about the Gluster-users mailing list