I am thinking of a time based index where we manage the lifecycle of a certain type of documents. The rollover API is one of the options vs spawning a new index at a cadence(say monthly). The nature of our documents are such that a document coming in for the first time will be stored in the latest index(indexes are by create date). These documents tend to get updated a few times over its lifecycle. Does the rollover pattern imply that any updates to documents in older docs would also go into the new/current index? Or is it possible to query for a document across all indices and update it in it's original index?
Into the current index, yes.
You can do that, but it's custom code, not native functionality.
Using rollover in use cases where you update documents can be trick and inefficient. It works best with immutable data. If you have a timestamp when the initial document was created and can access this when you update, you may be better off using standard time-based indices where each index covers a predictable time period. Thius allows you to know which index to update, which makes updates less costly, and also gives you time based indices that are easy and efficient to delete once retention period has passed.
If this is not the case I believe you either need to pair each update with a search (2 separate operations) when using rollover or not use time-based indices at all. The first option adds complexity and makes updates more expensive, but may work if the number of updates is low. The second option makes upodates easy but complicates index management and data deletion.
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.