A follow up: I've gone some steps further by manually moving some data (4GB out of a 1,4TB brick) and then issuing a rebalance again seems to do the rest of the job (already working 2 hours, moving data out of the full bricks to the others). So two new questions arise: 1. Is this procedure considered to be safe ? ( stop volume -> stop glusterd -> mv data between bricks -> start glusterd -> start volume) 2. Could the problem occur due to the fact that there was insufficient space for the .glusterfs directory to write data ? a. On 01/12/2012 11:54 ??, Alexandros Soumplis wrote: > I have a distributed gluster volume on 2 nodes. This volume had 1 > brick on each node and by time it became full. I have added 3 extra > bricks and issued a rebalance on the volume, yet one of the original > bricks is still 100% full and this causes the dht-translator to throw > disk full errors on clients when it attempts to write on this brick as > it seems it does not take into account the fact that this brick is full. > > The question is whether it is possible to manually move data from the > full brick to other bricks with plenty of available space and if > possible, how can I do it (I suppose it is not just a volume stop and > data mv) > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://supercolony.gluster.org/mailman/listinfo/gluster-users
