Is it possible to use this client on Heroku? about elastic4s HOT 5 CLOSED

Have a read of this
http://www.elasticsearch.org/guide/reference/modules/http/

And then the examples on elastic4s on how to use transports.

If that's not helpful then post up your config file and I'll help you further.

from elastic4s.

Hi Sam,

I am attempting to use a third-party elasticsearch hosting provider. The problem is they only offer connectivity over HTTP, and I have not been able to get this client to connect over that. (Not even to a local elasticsearch running on my machine.) I have tried to connect to a stand-alone elastic search instance over http by doing this:

val client = ElasticClient.remote("127.0.0.1", 9200)

However, this fails with a org.elasticsearch.client.transport.NoNodeAvailableException: No node available exception. I can absolutely connect to that over the REST API using CURL from the commandline, so the node is alive and well:

curl '127.0.0.1:9200/twitter/_search?q=foo'

It appears that the remote method assumes it should be making a TCP connection. I have tested this by changing the connection port to 9300, which is the default TCP port, and this works.

Unfortunately, as stated above, none of the third-party elastic search providers I've looked at exposes a TCP endpoint. And self-hosting is not an option for us at the moment, unfortunately.

This client has the nicest DSL of the clients I've looked at, so I would prefer to use it. But if it cannot work over HTTP I simply cannot at this time. :-(

Stig

from elastic4s.

Ok I understand now. Then no you can't use elastic4s or any of the other clients. Your only option is https://github.com/searchbox-io/Jest

I guess since I'm generating lots of json inside the client it would be possible to turn elastic4s into a full http client as well as a node client, I'll look at that.

from elastic4s.

As long as it doesn't compromise the backend :-)

from elastic4s.

In pyes, I'll alternate thrift and http on the same layer in python.
On JVM there are always pros/contros on using native client (port 9300).

Pros mainly are:

• faster
• high dependence from elasticsearch.jar, just is always updated

Contros:

• unable to use third part calls/plugin: they only work on http or thrift
• the internal ElasticSearch APIs change very fast. Requiring continuos migration of maintained source base.
• the elasticsearch.jar overhead ~10Mb plus others.
  I think that in future using several other protocols is a must.

The documentation elastic4s refers to HTTP calls, but we know that elastic4s now executes only native calls.

To provide HTTP calls without depending on elasticsearch.jar a lot of classes mapping queries, filters and facets must be written. I'll give a look to jest (last time I looked its code it was not well structured. I hate write JSON by hand. )
I'll create a branch to test the possibility to refactory the code to used both for native than for HTPP/Thrift. Probably we'll need to create a DSL factory that it's mapped to two implementation native and HTTP.

from elastic4s.