Cause of problem:

1. The reason for this problem is that the original slice was not properly closed and cleaned up, so there is no way to get a slice lock when the slice has to be reallocated back to the problem node.
2. This does not cause the slice data to be lost, it just needs to retrigger the allocation

Recovery instruction

curl -XPOST http://localhost:9200/_cluster/reroute?retry_failed

View details

curl -XGET http://localhost:9200/_cluster/allocation/explain