I have a cluster with two data nodes (d1, d2) and I have two indices (i1, i2) with 1 shard per each (no replica yet).
When I started to index my whole data (around 70GB) one of the primary shards were routed to the first and the other one was routed to the second node. After a while, d1 storage ran out of space so ElasticsSearch started to route my i1 shard to d2.
I managed to create enough free space to have i1 on d1 so I did a rerouting (from d2 to d1).
After it has finished, I still can i1 folder on d2 and I guess (not sure but it seems) that if I query that, it uses the d2 node.
Maybe I did something not properly but I guess after rerouting i1 from d2 to d1 I would expect that the folder is deleted.
What can I do now?