gluster volume disbalance

Posted on

we set up a distributed disperse gluster volume with such settings:
Type: Distributed-Disperse
Number of Bricks: 3 x (8 + 4) = 36
Transport-type: tcp
Options Reconfigured:
cluster.rebal-throttle: lazy
features.shard-block-size: 1024MB
features.shard: enable
disperse.background-heals: 18
cluster.background-self-heal-count: 20
cluster.disperse-self-heal-daemon: enable
cluster.heal-timeout: 500
disperse.self-heal-window-size: 4
cluster.data-self-heal-algorithm: diff
cluster.self-heal-readdir-size: 2KB
cluster.self-heal-window-size: 2
disperse.shd-max-threads: 6
storage.build-pgfid: off
storage.owner-gid: 36
storage.owner-uid: 36
performance.global-cache-invalidation: false
transport.address-family: inet
nfs.disable: on
36 bricks on 6 nodes.
12 bricks (2 on each nodes) out of 36 have half the free space than the rest.There are also twice as many shards on these bricks. Rebalancing didn't help to align used space on the bricks.
Could you advise how to fix the situation.
There is also the question: what will happen when the free space on these bricks runs out.