We heve Elasticsearch cluster setup with 3 client nodes and 15 data nodes. Index rotation happens once a day and something went wrong while indexes had been rotated. Disk IO has become 100% and was occupated by Elasticsearch. Also size of indexes started to grow extreamly fast with SAME amount of records. I had no posibility to make an requests to cluster due to reqest timeout. This behavior was fixed after cluster reboot.
We are using ES 6.2.2