Get rid of unassigned shard
I actually ran into this situation with ElasticSearch 1.5 just the other day. After initially getting the same error, I simply repeated the /_cluster/reroute request the next day for lack of other ideas, and it worked, and it put the cluster back into a green state immediately.