Geo-replication techniques?

From what I understand ElasticSearch does not offer any features to do geo-replication. How are people addressing this issue for continuous data feed to ElasticSearch, async, with say 10s delay? I had a 1 year long break from ElasticSearch. I remember there were some talks about Tribes.
Some possibilities I can think of:

  • Populate ES cluster at each location separately from some geo-replicated queue?
  • Tail and replicate transaction log (Translog) – is it possible, how?
  • River?
  • Snapshot?
  • Can I use Tribes for this?
  • Other?

What is recommended or works in production?

Check out