It has been a while (I think) since this question came up on here, so I thought I'd check whether the answer had changed or not.
I have a transform which produces consolidated data with a time-series element to it.
This data is currently being written out to one index, but since the volume of the data is significant and the retention requirement lengthy, I would like to roll the index over from time to time.
Ideally, I'd like to use a datastream to handle this but, historically, this has not been possible.
The next-best alternative answer I could find was to use an index with an ingest pipeline which uses the
date_index_name processor (as documented here) and a standard ILM policy on the destination indices to handle tiering, etc.
Is the date_index_name method still the recommended method?