[flexer2]

We use ES heavily. Most of our queries are basic document filtering plus some geospatial stuff. We could probably have done it with Postgres/PostGIS, but with AWS manages ES, it's all "good enough" -- we can do geospatial searches on millions of documents with response times around 100ms. The other part I like about ES is how it's easy to scale out across machines, which lets us handle quite a bit of load and tolerate failures easily. We have a cluster of 5 m4.large instances and it only runs us about $600/mo. Like others have said, tuning AWS ES sucks, but it's always been good enough for us.

We've run into some pain points like trying to index very large shapes into a geospatial index, but have workarounds for basically everything now. We also had a problem where when AWS had the outage around autoscaling groups a few months ago, we lost 3/5 of our instances and had to reindex some data from backups. That was the worst thing that's happened.

I'm sure there would be better/faster/cheaper ways of doing what we do, but for what we get out of the box for the price, it's going to take a lot for us to move away from it for now.