I have a collection of around 100,000,000 documents, which consists of a few thousand subsets (each tens to millions of documents). Most of these subsets are regularly updated (wipe and replace, a few times per month). Until now my strategy has been to first remove the old version of a subset and then index the newer version, but this means that during indexing (which can take up to half an hour) the subset is not or only partially available. Would it be a better strategy to create an index for each subset and just update the alias after a newer version of a subset has been indexed? Any caveats?
Thanks