How to create a new ElasticsearchHttpClient class implementing org.elasticache.client.Client to enable ES applications to run against AWS Elasticsearch?

I am trying to add Elasticsearch HTTP access to the Titan ES client using JEST. titan-es only supports ES' local and transport (TCP) mode. But I would like to support communication over ES' HTTP interface. That would allow client libraries like titan-es to use AWS Elasticsearch as an indexing backend which only provides a HTTP(S) interface. See this post for more information.

I am looking for some feedback on the approach I am considering so far:

  1. Create a new class ElasticsearchHttpClient that implements the org.elasticache.client.Client interface. The new class will use the JestClient as it's internal client. This way it will communicate with ES over HTTP. The new class will likely extend ES' AbstractClient to reduce the methods that have to be implemented to: admin(), settings(), execute(), threadPool(), and close().
  2. Add a new enum HTTP_CLIENT to ElasticSearchSetup
  3. Ensure that the connect() method on HTTP_CLIENT returns an instance of Connection which contains proper values for node and client. The client member would be an instance of the new ElasticsearchHttpClient class.
  4. Ensure that ElasticSearchIndex.interfaceConfiguration() method retrieves the correct instance of Connection (containing the new ElasticsearchHttpClient) if the INTERFACE is configured as HTTP_CLIENT. From that point on the rest of the code should continue to work over the new protocol.

Does that sound like it should work? The 1st step is my biggest concern - I am not confident that I can implement all Client methods using the JestClient.

Has somebody tried this before?


This was answered on the Titan mailing list.