Data Re-Indexing

Every so often you may run into changes required in mapping for existing fields
As per our previous attempts updating mapping for existing fields is not possible
The only solution I am aware of: create a new index with new mapping and move data over
To move data reliably in ETL (extract, transform, and load) like fashion: scan API is useful

To initiate a scroll:

curl 'localhost:9200/logstash-2015.05.20/_search?scroll=1m'

To continue with scroll until no documents left:

curl -XGET  'localhost:9200/_search/scroll'  -d '
{
  "scroll" : "1m", 
  "scroll_id" : "c2Nhbjs2OzM0NDg1ODpzRlBLc0FXNlNyNm5JWUc1" 
}
'

What do you do with the data you receive?
How do you plan for re-indexing possibility without affecting cluster's clients?