Hello everyone,
We are building a new cluster OS 3.2 from an old ES1. We work on a multi-tenant app.
We don’t want any risk of mixing data between instances. We have 7 different types of contents (like articles, events, jobs, companies, profile members, feed post…) indexed today.
We used to have an index for each instance, around 350 index. For this new implementation, we are wondering what could be the more interesting/efficient with OS 3.2 :
- one index per instance (350 index) like we did before
- or one index per type of data per instance (350*7=2450 indexes) - risk of shards issue ?
- or one index per type of data for all instances (7 indexes, easier to manage but with a risk of mixed data )
We are also planning to make semantics search for some specific data in the near future.
Any advice will be appreciated 
Thanks