Let me answer your point, ES sells itself on being a fast text based search engine, running on the NoSQL principle. All well and good, but the documentation is online, unorganised in as much as start here and watch out for some simple mistakes. The latter has cost me a huge amount of time and since time is money, money. NoSQL has as a principle that no schema is need, unlike a RDBMS, unless I'm mistaken. Then it appears that I do not have to define my document structure beforehand. So, I load up my data which takes the best part of 13+ hours and that's only a subset of my full dataset. To load the data as quickly as possible, since it's streaming of a pseudo real-time system, I do not define complicated tokenisers. Then I go to add a whitespace tokeniser to a field, only to find out that I can't do so without reindexing the entire dataset which takes a similar amount of time. I then for reasons of time need to get at one slice of data, identified by a numeric value, and find I'm extracting around 1000 records every five seconds which is ludicrous given I'm intending to extract 6.2 million or thereabouts. I am then told that indexes on numeric feeds are inefficient, and I need to change the field to a keyword and that will require reindexing the data again.
ES is to be admired for what it does well i.e. inexact text indexing, and that is necessary for what I need, but these issues have made the task of loading the data and indexing it as I need it to be well nigh impossible.
The documentation which seems initially to be thorough needs a serious look at. It seems, after the fact, to be more a stream of consciousness or a single stream of thought assuming there reader will start at the beginning and read all the way through with the same level of comprehension.
In summary, I should have been able to do simple things simply. Load up a single stream of records of fixed number of fields and types into a single index, and queried them. Instead, I've found myself spending hours trying to work out how to do simple things. Even the text books seem to fail to point out such simple 'gotchas'.