Elasticsearch as a data source

In general, Elasticsearch shouldn’t (subjective, yes we do acknowledge this) be used as a primary data store. However, this question is more use case-driven and for some use cases it could very well be used as a data store. Elasticsearch does fall into the NoSQL type of database and doesn't support the ACID property of a typical relational data store, mostly used for transaction-oriented use cases. But it does have features such as optimistic locking and eventual consistency making it apt for certain pointed use cases. For a data lake implementation, it could very well act as a data store because the real data store (system of record) is with the source systems. In the case of any failure, the data could very well be warmed into Elasticsearch (in practical scenarios this is not that straight forward.. smiley) from these source system or even from our Hadoop and back to...