I have many nodes (version 1.7.1), and I disabled shard shard balancing:
I then move a shard from one node to another using
/_cluster/reroute, which is successful.
The problem is the data for the shard remains on the original node
After many moves, both in and out of a node, and across many different indexes, the drive fills up. I must login to the node, and delete the directories for shards that have long left that node to free up space.
Is there a setting that will remove data from disk for shards that do not exist?