threshold not working properly?

[dwatzke]

Hello!

I don't know if I understand the threshold option correctly, but I consider this message bogus:

[root@ams02-008 ~]# /opt/volume-balancer/bin/cdh-balance-local-volumes.sh -threshold=0.15 [...] 14/10/21 11:07:08 INFO datanode.VolumeBalancer: Least used volumes is within the threshold, we can stop.

when there are disks that have almost no free space (/data/05 and /data/08) and others that are at half their capacity:

Filesystem 1K-blocks Used Available Use% Mounted on /dev/sdb 960906608 445983476 505160892 47% /data/01 /dev/sdc 960906608 449322924 501821444 48% /data/02 /dev/sdd 960906608 447735880 503408488 48% /data/03 /dev/sde 960906608 603740396 347403972 64% /data/04 /dev/sdf 960906608 942002824 9141544 100% /data/05 /dev/sdg 960906608 412422180 538722188 44% /data/06 /dev/sdh 960906608 452300956 498843412 48% /data/07 /dev/sdi 960906608 938984604 12159764 99% /data/08 /dev/sdj 960906608 411759028 539385340 44% /data/09 /dev/sdk 960906608 424126144 527018224 45% /data/10 /dev/sdv 960906608 454352928 496791440 48% /data/21 /dev/sdw 960906608 650886156 300258212 69% /data/22 /dev/sdx 960906608 446865056 504279312 47% /data/23 /dev/sdy 960906608 420681332 530463036 45% /data/24

The difference between those disks and for example /data/06 and /data/09 is huge, way bigger than 15%.

Can you please look into this and possibly make it a priority for the tool to free some space on fully utilized disks first, before anything else?

Thanks!