i want to put a 1:1 elastic cluster inside my hadoop cluster, 1 elastic node on each hadoop datanode.

to not interfere too much with the hadoop cluster, i would like to run the elastic nodes on a disk of it's own.

the setup:

CDH 5.6.0

datanode disk layout(in mounted dirs):

/data/disk1

.

.

.

/data/disk10

20+ data nodes

say i wish to remove disk10 from each datanode, how do i do that without data loss?

removing the disk on a decommissioned datanote, and later recomissioning it, takes too long time. - any hint on making this process faster?

can i use the rebalancer? ( i saw there is internal datanode balancer in CDH5.8)